A conversation with Mohit on credit investing and AI

Welcome to the 217th Pari Passu Edition.  

In addition to our usual Friday deep dive on Fortrex (Packers Sanitation Services), this week we are publishing my conversation with Mohit Hajarnis.

Over the last few years, many of you have asked for a podcast and more restructuring-focused interviews. Research is our focus, so we have generally pushed back. Today, we are making an exception.

Mohit sits at the intersection of credit investing and AI, and I have learned a lot from our private conversations. He started his career as an analyst in Goldman Sachs’s Special Situations Group, spent three years at Diameter, and then three more as a senior analyst at Citadel. In 2025, he left to found Constellation, a venture-backed company that’s building what we think is the next category in Financial Services AI.

Whether you are a banking analyst trying to break into credit, a single-manager analyst weighing a move to a multi-manager platform, or an investor or banker trying to figure out what’s next in AI, I am confident you will find this conversation valuable.

Let’s get to it!

Pari Passu: Before getting started, I would like you to fully introduce yourself. Could you give me a 90-second background on your background path and what you are working on today?

Mohit: Awesome. Well, good to be here with you. A little bit about me. I started my career in 2016 at Goldman. I was in SSG there. I did that for 3.5 years. I moved to what was then a small hedge fund called Diameter, now a very large hedge fund called Diameter. I was there for another three years, roughly. And then I got recruited to go to Citadel, where I covered industrials and energy for one of the high-yield credit books. And I worked at Citadel for a little over two years. And then in the summer of 2025, I made the decision to leave and start Constellation. At the time, Constellation was focused on solving problems specifically in credit, for analysts and lawyers. And then we've since expanded into a more ambitious version of that, helping all sorts of financial services companies and law firms solve the very minute problems that they have every day, using AI. 

Pari Passu: Perfect. And we'll dive into Constellation and AI later, but a lot of our audience is bankers, investors, and lawyers, so it would be very good to hear about your experience at the investing firm. Starting with Diameter, obviously you have witnessed a very large expansion of the fund. It would be great to hear how the investment process evolved. I assume maybe the book was just central and working directly with the PMs, and maybe over time it got layered. It would be great to hear how your experience evolved over your time there.

Mohit: The way Diameter's decision-making works is that Jon [Jonathan Lewinsohn] and Scott [Scott Goodwin] are obviously the co-PMs there. Scott's more focused on the trading side. Jon's focused on research. And it's always been that way, and I think one of the great things about Diameter is that even as it got bigger, it stayed that way, that they were the key decision makers. They now have a head of research, Eugene, and people running other parts of the business like CLOs and private credit. But at the end of the day, in the master fund, it's really Jon and Scott still making decisions. And I think it's really important because a lot of what makes Diameter special is the two of them. And so keeping their role in the firm, and even as the firm has grown, not diluting that, has been, I think, a key reason for its success.

Pari Passu: And in terms of your responsibility, were you a generalist at Diameter or did you have a specific sector coverage?

Mohit: When I joined Diameter in July of 2019, I was covering transportation and energy services. And I covered that through the balance of COVID. In 2022, I picked up gaming. So I was focused on subsets of this. When I went over to Citadel, part of the reason for doing that was being able to cover, albeit with a much smaller team, being able to cover more of the high-yield universe.

Pari Passu: Got it. Can you give us more of a primer on how credit hedge funds think about their book? It would be great if you could give a two-minute primer on how people at Diameter thought about it, and maybe how you thought about the book structure at Citadel.

Mohit: I think every hedge fund is a little unique. I think the way that Diameter thought about book construction was, one, we ran over 100% gross and 60-70% net, which is in line with most single managers. Shorting was a big part of the book at Diameter. They wanted to run more of an alpha-generating strategy. The second thing is that they thought about it in terms of how to play more macro themes in the credit markets through individual positions. 

So on a given day it could be, what's the best or the worst thing in BB? What's the best or the worst thing in this industry? And layered on top of that, that's sort of the trading angle to it, and layered on top of that is, how do you capture these microcycles in specific sub-industries, whether it's aircraft leasing or chemicals or a particular part of the energy complex. So that's really how they thought about it. And away from that, what was really great about being in that structure was that it was relatively unconstrained. So we did everything from investment grade through deep distress, process work. It ran the gamut of what was out there.

Pari Passu: Can you elaborate on that, maybe provide an example of how you would play the best theme in BB?

Mohit: So you can say, “I've got a theme today. CCCs are in favor because the market is looking to get more risk on.” So it's a trading read on what people are willing to buy and sell. The reason this matters is because the large institutional investors, the BlackRock and Fidelity of the world, allocate massive amounts of money in and out of these more thematic ideas. So they want to say, “I think that there's a lot of excess spread today in BBs relative to the average market. So if I'm trying to beat high yield, which has all of these things in it, I'm going to go buy some more BBs.” Or “I think CCCs are getting a little long in the tooth, so I'm going to sell CCCs.” 

So Diameter thinks about the book a little bit more thematically. Another example could be along the lines of: “It's a good time to be in the high-grade part of BB, so we want to be heavier in BB+. Sector analysts, if you had to go buy more high-quality BB names in your sector, which names would you buy? 

So that's how that idea was getting expressed. It was expressing a market view through individual securities.

Pari Passu: This is very interesting. What are some other examples of trades or positioning or setups that were a recurring theme for how credit hedge funds think about their books?

Mohit: The number one way that we were taught to invest was looking for industries, sub-industries, or parts of industries that we covered that were going through periods of change. So one of the things that we would do is at the end of every year, we would have this Diameter offsite, and at the offsite, we'd have to present a piece of primary source research. So something that we had gone out and researched ourselves. And then use that to inform trading views or inform positions for the following year. An example of this was in 2019, my senior analyst and I presented on the U.S. airline cycle, and why we felt like the airline cycle was a little bit long in the tooth. And we used that to inform our positioning for, at that point, it was the next three months, but it influenced how we thought about airlines for the next two years.

Pari Passu: So is it fair to say the PM has a top-down view of the market and then he would ask the fundamental analyst what is the best name in your coverage that fits this criteria, and then the two blend together and form a trading strategy? Do you think it's a fair characterization of the model?

Mohit: Yeah, that's exactly right. Scott and the traders had a really good pulse on the market and an understanding of what people were excited about buying and selling, and then they would come to the research team to say, okay, how can we express that in individual securities?

Pari Passu: What do you think about the overall assumption at the base of this approach, are you trying to front-run the real-money [BlackRock, Fidelity, etc.] guys?

Mohit: I think the right way to think about it is, in high yield, particularly, you're trying to figure out who is going to buy this bond from me and for what reason. And so in that theme, you're trying to figure out what it is that people are excited about buying and selling. That's part of the strategy, trying to figure out what it is that people actually want to buy and sell on a given day.

Pari Passu: And how quickly could it take to flip in and out of a name? 

Mohit: It could be as little as one to two days or as long as a quarter. It's really a function of how long that capital flow persists and what the drivers behind it are.

Pari Passu: Got it. But even for the thematic, industry bottoms-up trades, can you get in and out that quickly? The fundamentals don't change in days. I would assume an industry thesis has to have a base of fundamental inflection?

Mohit: For something fundamental, it could be as easy as, “I think numbers are going to inflect this quarter, and it's a trade for a couple of days or weeks around earnings.” Or it could be as long as, “I think the numbers are going to get better over the next two quarters, or this is a one-year story.” What was nice about Diameter is you had the flexibility to do a little bit of everything. So that was definitely, I think, a strength of the model.

Pari Passu: Got it. And maybe can you talk a little bit more about position sizing? You talked about gross and net, but how many positions would go into that? Let's say you were doing this type of CCC trade. How would you size it, how many positions would go in that trade, and what position size would it be?

Mohit: It could vary very much based on the analyst's level of conviction, the PM's level of conviction, the trader's level of conviction, and whether the technical was supportive or not. The analyst's track record and how much credibility they have, also mattered into the equation. I would say, Diameter ran a book that was probably on the order of 200 positions. So everything was roughly a 50 basis point position, but there were things that were really big. Diameter was in a number of COVID bankruptcies or pre-COVID bankruptcies that ended up being quite large, like PG&E, for example. So that was a larger position. And then some stuff could be very small, or slightly more off-the-run.

Pari Passu: Got it. So would it be fair to characterize a 25bps as a small position and a 250bps as a large one? 

Mohit: Yeah.

Pari Passu: Got it. And can you maybe blend this portfolio sizing point with liquidity? Let's say you have a 50 basis point position; that's your target. What was the target capital structure size that you would look at? How would it blend? Let's assume you have a one billion dollar book, and you want to have a 50 basis point position. How quickly do you want to be able to get in and out? How did you think about liquidity in general?

Mohit: I don't think there were any hard and fast rules, but you always had to be mindful of the size of the capital structure and the relative liquidity of any given position. I would say that most of the stuff Diameter did was more weighted towards more liquid things. That was an advantage for when either the market was changing or the fundamentals were changing or you just felt like you had realized the price target, and you could move in and out of a position quickly. 

This is very different from some credit hedge funds where they specifically traffic in very illiquid things, and they want to be able to either own a lot of it, influence the process, or move the price. Diameter took the opposite approach. 

Pari Passu: Got it. So basically a mix of both, but it wasn't a defined "we need to get out in a week." There wasn't that type of structure.

Mohit: No, there's nothing super defined about it.

Pari Passu: Great, and to finish on this theme of portfolio construction, most would agree Diameter has been a pioneer playing the same name on both the short and long side [eg. short a bond at 100s, cover at 80s, and long once it hits 70c]. Can you talk a little bit more about the model? 

Mohit: Diameter requires you to be very intellectually honest. If you take a step back, Diameter's roots are really in the trading desks. And so they were very comfortable flipping between long and short in the same name. They're very comfortable being fast and tactical. And so that's very different from how a lot of single managers invest, where they're typically just long, or rarely just short. So part of doing the job at Diameter was being able to be intellectually honest about when the thesis had changed, when the fundamentals had changed, when there were things that were working in favor of or against your position. And really, when fair value had separated from price. And I think that's a really difficult skill that the Diameter team did really well.

Pari Passu: Got it. It's very helpful. Let's say you're new, you work in banking or in PE, you join Diameter, and you're given a name. Over time, what did your process evolve to and what was the progression to getting up to speed, trying to achieve differentiation versus the market, and getting a name into the book? Was this an initial meeting with Scott and Jon, or maybe one with the traders?

Mohit: I would say that getting up to speed is no different than how most other hedge funds do it. Every analyst had their own process. Some people relied more heavily on expert calls to get up to speed; some people relied more on the sell side. If it was a new issue, then you would talk to the market, or to management, to get up to speed. People relied a lot on their relationships to talk to other hedge funds to understand how they thought about the name. And then once you have formed a coherent view, what was unique about Diameter is it was a very no-memo environment, so it would typically be an email sent to Jon and Scott, and then you would typically get on a Zoom together and walk through your model, or your email, live, while you were on the phone with them. The process has changed now. You go through the head of research, or one of the heads of research, and then you go to Jon and Scott. But during my time there, it was a pretty flat organizational structure. You spoke to them directly on those things.

Pari Passu: How did you think about allocating your time? Let's say there's a name that you've never heard of, and you're given it. Do you think about a certain number of days, a certain number of hours? Over time, since you did this for a lot of years, how do you think about becoming very efficient, not wasting time, while at the same time achieving the differentiation?

Mohit: I would say that early in your career, you're just going to have to spend a certain amount of time on something. You're going to have to spend a day, two days, or a week. That's the amount of work you can do at a place like Diameter, maybe a month or a bit longer. But the reality is that as you get more senior and more tenured, you just have more pattern recognition. And you realize, okay, after doing a couple of hours of work, I know that this smells a lot like something I've looked at before, something that's made money before. My instinct tells me that this is a good trade. Versus before, where you had to get through all of the work and then try to figure that out.

Pari Passu: Especially as people think about doing case studies for hedge funds or doing pitches, what are some good tips that you learned about pitching a name on either the long or the short side? And on the same or opposite side, what are some common mistakes that maybe you or analysts made when pitching to Jon and Scott?

Mohit: A very easy thing that separates good analysts from great analysts, or good analysts from bad analysts, is how you think about and approach each key line in your model. Because part of doing a good job as an analyst is understanding the business model and the drivers of the business. So one of the things that we learned really early on, which I've kept throughout my career, is if you're a banking analyst, you're often taught to model gross margin, just percentage margin. But gross profit margin is an output of gross profit, which is itself an output of fixed and variable costs and revenue. 

So key question is #1 is: are you able to model the business's fixed costs and variable costs? Do you understand what the drivers of variable costs are? Do you understand how commodities might factor into that, or how energy might factor into that, etc. 

Companies typically give you a decent amount of clarity on these things in their financials. And so being able to go through that and put it together is really important. And that's really a differentiating factor, I think, for good analysts versus great analysts. And so that was a big part of what people did well versus what people did poorly, especially when we look at case studies: how much detail, or how thoughtfully, did they model something? And thoughtfulness is not always the most granular level of detail, like did you get every single COGS item? It's more about: did you understand what's actually driving this business, and why numbers are going to go one way or another?

Pari Passu: What about the part about pitching or differentiation? How would you go about having a very specific answer to why we are different from the market, and when the gap is going to close, and we're going to make money?

Mohit: I think understanding your catalyst is obviously a big part of the trade. Are you waiting for earnings? Are you waiting for something specific? Because as I said, the fundamental question that you're trying to answer always is, why is somebody going to buy this bond or loan from me? Or why is somebody going to sell this bond or loan back to me if I've shorted it? A big part of that is understanding what are the circumstances that need to change in order for that to happen. 

Let me give two examples. 

We were doing the LATAM bridge deal. For context, LATAM Airlines is the largest airline in South America. It had filed for bankruptcy in May of 2020 and was preparing to exit bankruptcy in early 2022. To fund the bankruptcy exit and repay the DIP, the company had come to market with a term loan and secured bond, but it came to market at a slightly challenging time around Russia-Ukraine. The issue was that the company had to exit bankruptcy regardless of market conditions, so some of the larger lenders like us, HBK, and a few others stepped up and backstopped the exit financing with JP Morgan. For whatever reason, the market wasn’t buying the deal, so we got hung with a lot of the risk that couldn’t get syndicated. We had said, all right, we think that these things have to change in the market; this has to get proven out, in order for people to want to buy this away from us. Once those events were realized, we sold our position got sold. 

Another good example is Rolls-Royce CDS. We sold a bunch of Rolls-Royce CDS in early/mid '22, when it had blown out because of Russia-Ukraine. And we said, what needs to change? Once Rolls-Royce goes investment grade, which the ratings agency said it would at the end of this year, then CVA [Credit Valuation Adjustment] desks and banks are going to sell CDS, because they no longer have to hedge against it. There's a structural reason for this happening. 

So being able to understand the fundamental drivers, or why something is going to change, and when it is going to change, is really important here.

Constellation is opening a small, invitation-only beta for investment firms with fewer than five investment professionals that want more capacity without adding headcount.

If you have more ideas than hours to chase them, and you’ve wished for one more analyst on the team, this is for you.

Beta participants will work (for free) directly with Mohit and the Constellation Research team to shape what comes next.

Interested? Reach out to [email protected], and we’ll make the introduction, or directly from their website. Spots are limited.

Pari Passu: Just to transition a little bit, a very common path that people take is the classic banking, into single manager, to multi-manager. And obviously you followed that too. Can you maybe talk a little bit more about what's the structural challenge of a single manager that leads so many analysts, especially successful ones, to go down this path?

Mohit: I think it's less about what leads analysts to go to multi-managers and more about what leads PMs to go to multi-managers. If you're a very good senior analyst at a hedge fund, you may choose to go to a multi-manager because there are a few PMs, or at a single manager, all the decisions are made by one or two people. And so you want the opportunity to put your name on the door and make money, or make more money, based on your actual outcome. Because if you think about how much you're paid at a multi-manager relative to your P&L, it's much higher than what you're paid at a single manager relative to your P&L. But that can mean a lot of different things in a lot of different places. And so part of the reason why people leave is more autonomy, more independence, and then ultimately compensation.

Pari Passu: Got it. And to the extent you're comfortable sharing, if you are an analyst and spend a few years in banking, then a single manager, and you move to a multi-manager, what kinds of things were you able to negotiate? What advice would you give to someone going through the process? What are the key things they should be focused on getting when they are talking with either the PM they're going to work for or the business development person at a pod?

Mohit: I think the questions that you're going to have are, primarily, what's my compensation, and what sectors am I covering, or subsectors? That's a discussion with your PM. And one of the things about multi-managers is that very little is negotiable beyond compensation. So you're going to get the deal that you get; you're going to have the non-compete or the sit-out that you have. There's not much you can do about changing that. But it's really just how you can set yourself up for the most success, which is mostly around expectations. 

“What P&L do you expect me to produce in the first couple of months or weeks, or a year?” 

Needless to say, a huge part of being a multi-manager is that you're expected to produce P&L. There isn't much of an on-ramp period. And so giving yourself that grace and understanding what other people's expectations are is a really important part of the process.

Pari Passu: Let's say a friend comes to you asking for advice. He is a single manager analyst who thinks he's very good, and so he is leaving for a multi-manager. He has done all the case studies, and they want to hire him, and now he has to finalize his contract. 

What's the specific advice you would give? What's very tactical advice that you wish you could tell a single manager analyst who was moving to a multi-manager role, to try to get a better deal and negotiate?

Mohit: I think that goes back to, when you're going to a multi-manager, there's not much negotiation. Citadel hires hundreds of analysts a year and fires hundreds of analysts a year. And that's true for Millennium and Balyasny and everybody else too, so that's not Citadel-specific. So the reality is really just: negotiate your compensation well, get the most money that you can get. And then the rest of it is up to the PM and up to the business, on how they want to, what they expect from you. But it's very difficult to get much beyond that in writing. And so the reality is that you're really at the whim of the process there. So you can really just set realistic expectations for yourself and then try to maximize compensation.

Pari Passu: In terms of seniority, what do you usually see? Were you hired as a senior analyst? What does a senior analyst at a multi-manager really mean? What were your responsibility expectations?

Mohit: It's a good question. So a senior analyst at a multi-manager is somebody who can operate independently and make decisions for themselves, can cover a sector well, because typically these are sector coverage roles, can basically recommend trades, generate alpha in that sector, and then be in a position to start to manage their own book after some period of time. So I think that people should think about transitioning over to the seat when they feel like they can cover a sector really well, where they're not missing a lot of things, where they're generating alpha relative to their sector. And then they're sort of ready to be on the path to managing their own book over some period of time.

Pari Passu: So is the right way to think about it that the first moment you move, you're effectively still going to be an analyst, in the sense that you don't put in the trade? You go to the PM and say: we should buy this, and hopefully the PM trusts you, and then hopefully over time you get a carve-out? What's the next step?

Mohit: I think that's right. Different firms do it differently. Some firms would carve out, some firms have an associate PM position. There are a lot of different ways to do it, but yes, you want to be on some sort of path to eventually being a PM. And some firms, some groups within those firms, have more explicit paths of, here's what our analyst-to-PM pipeline looks like, here's how we cultivate PMs, here's how we teach PMs, et cetera. So it varies in that sense.

Pari Passu: Is there anything that you did when you went from analyst at a single manager that helped you develop that skill of portfolio manager? Because a very common pushback against the switch is that it's a completely different job. Being an analyst, covering name to name, is very different from thinking about a sector, and thinking about the book. How did you adjust, and what were some of the biggest learnings from there?

Mohit: I would say the biggest learning of all, and this isn't necessarily even just single manager versus multi-manager, is that the ways in which each fund or firm or PM makes money are very different. Trades that certain PMs would do, other PMs wouldn't do, even if they both knew that those trades were going to be profitable. PMs have styles, they have preferences, things that they've been burned on in the past. They have sectors that they won't touch. They have sectors that they like a lot, even too much. So all of these things matter.

So part of finding an analyst seat is finding a seat where your process for making money fits into that seat. If I'm somebody that only does long-only distress, I'm not going to be very successful in a seat where I'm expected to run a balanced book. Or if I'm somebody that only takes $5 to $10 million positions and runs a factor-neutral book, I may not be that successful at a seat where I'm expected to take large swings. 

While the first analyst seat should be focused on maximizing learning, for subsequent ones, my advice would be to really focus on finding a seat where you can maximize P&L in the same style that you are good at.

Pari Passu: Very helpful. And digging deeper, you mentioned earlier that you should think about delivering a certain amount of P&L, dollar P&L. Can you walk us through some basic math starting from $100 of investor capital and bridging us to dollars of P&L in a multi-manager context? 

Mohit: The platform raises $100 of investor capital. The platform runs the capital at some level of gross leverage, we can assume 5x, to get to $500 of GMV [gross market value]. The GMV is allocated across pods, for simplicity, we can assume 5 pods, each receiving $100 of GMV. 

Together with GMV, each pod will receive a budget of P&L, a target figure. Let's say, $10 of P&L on that, which represents 10% of GMV. 

So the way that you get there is you say, okay, I'm going to be right 50% of the time. On my net winners minus losers, I'm going to be up four points. So that's 50% times four, so that's two points, or 2%. Since I have a target of $10 of P&L and my expected return is 2%, I need to make $500 of bets to achieve my target. Since my GMV is $100, I need to turn the book 5 times over the year to achieve my target. 

The last factor is position size. To illustrate, if I wanted to, I could keep all my capital in one position at a time. Otherwise, I can split it across 25 positions with an average size of $4. Note that I still need to turn the book five times, so if I have 25 positions at any time, I need 125 positions for the year. 

This is realistic for a very high-velocity desk, but that's not realistic for a desk where you have to take big swings infrequently with a higher target spread. 

There are firms that will put on five positions in a year. But at a multi-manager, your expectation is that you're going to run this process much more frequently. And that 125 positions can include longs and shorts and switching back and forth between things. But you're expecting to operate at a much higher frequency.

Pari Passu: Super helpful. And maybe just for context, can you elaborate on how a P&L budget is reached? Is this something that the firm gives you, is this something the PM gives the analyst, or the PM is given, because those are two reported numbers?

Mohit: It's a function of how much risk your PM gets from the firm and then your PM negotiating with the firm on what they think they can generate with that risk. So they may say, all right, if you give me more risk, I can generate more dollars of alpha out of it, or I'll promise a higher return on it than if you give me less risk, because if I have more risk, I can do all these other things that I couldn't have done otherwise. So that's the trade-off that you and your PM, or the PM and the firm, are always making.

Pari Passu: Got it. So effectively, the PM negotiates a gross market value figure, and based on that, he gets a dollar P&L target, and you take the dollar P&L target, and based on your success rate and spread, you arrive at a portfolio construction, and from there you need fundamental ideas to go in.

Mohit: Correct.

Pari Passu: Okay, perfect. On a similar topic, it is common to say that multi-managers pay very well if things are aligned, as there's a much bigger pie of P&L shared with investment professionals. Can you give a very short primer about pod economics? How does the pod, at the pod level, make money? And then how do you, as a senior analyst and then as a PM, make money?

Mohit: So the way pod economics work is, you have a certain “slope”, so that's the amount that your pod makes as a percentage of its P&L. And that can vary really widely, and not all slopes are made the same, because not all P&L is made the same. Some places calculate P&L only as alpha, or above a certain benchmark, or above a certain budget, or above a certain cost threshold. So people do it very differently. So number one is you're going to have some sort of P&L number that's affected by all these things. And then you have a slope against that number. And then, of what's remaining, you're typically required to take out desk expenses, although sometimes they can be in the P&L number itself, but you can have desk expenses. And then you've got a remaining pool. And of that, the PM gets to discretionarily decide how much to pay each analyst. And then the PM gets to keep the remainder. That's the math of how this works. And what's important is to understand that some firms will give the analyst a very specific share of, here's how much we will guarantee you as a percentage of slope. Some other places will just leave it up to the PM. In most cases, you'll just end up with a handshake agreement with the PM themselves. So that's what people need to be mindful of.

Pari Passu: Extremely helpful. And the natural follow-up is that obviously the percentage of profit is much higher than at a single manager, but you have a lot more constraints. And obviously the first constraint is how much beta you can run. But can you help us understand specifically how these constraints shape the way you deploy capital? Are you literally taking positions off if you're above certain factors? How does this constraint actually show up in the way you invest?

Mohit: It is very firm-specific. Typically, you have a risk limit; you can't cross the risk limit, or somebody from the risk department will come and tell you to get back within your risk limit. It's very much like working on a trading desk. And so that constrains what you can buy and sell at any given time, and it requires you to be thoughtful about how you're constructing your book. So you're always trying to figure out what is my highest-probability way to make money given the constraints that I have. So if I have a certain amount of beta that I can put into the book, you're trying to figure out what is the maximum unit of dollars of P&L per unit of beta that I have in my opportunity set, and that's where I want to put more money.

Pari Passu: Got it. Can you give more examples beyond beta of these constraints? Beta's kind of the obvious one that even a single manager analyst gets. What are the actual other constraints that limited your ability to make money?

Mohit: Every firm has different risk parameters and risk models. So some firms will use different individual factors, factors like big versus small, momentum, etc.. Some firms have custom factors or shocks that they use. So every firm is very different, but one of the things that people, in particular people that are going to PM seats, need to think about is, can they operate within those risk parameters? And are those risk parameters something that they're able to make money within?

Pari Passu: How do you develop that recognition when you're an analyst going idea through idea? Single-manager analysts, at least in my experience, don't often think about factors, for the most part. You are telling them to figure out if you can make money in a factor model. It seems like a very hard transition. Is there any way to smooth the process, or is it really that you make the jump and hopefully you're smart enough to figure it out quickly?

Mohit: I think even in a single manager seat, you have to think about, sometimes, why things are performing the way that they are. Why is this bond rallying, or why is this bond selling off, or this term loan, or this equity? And you realize that it can be a lot of different things at once. And once you put names to those forces, that's really what factors are. Why is this bond selling off? Because other consumer durables are selling off. Or why is this rallying today? Because BBs are rallying, because there are a lot of inflows. Or why is this selling off? Because rates are widening. So that's the same thing as factors, really, just understanding the drivers behind why something is moving. And I think what good analysts and great analysts can do in single manager seats is understand that things move for reasons other than fundamentals, and understanding what the factors of those moves are, understanding why those moves are happening, is then setting yourself up well to operate in a factor model.

Pari Passu: Very helpful, thank you for sharing. Is there anything else about the pod model that you think is underappreciated? The transition is something a lot of people discuss, but it's very hard to actually get clarity. Anything else that you would like to share about how you adjusted, tips on making sure it's the right idea, because for many people, being a career single manager is a very good path. Anything you'd like to share?

Mohit: I think the pod model works really well when it works well. People obviously do very well in it. It's going to be lucrative for the people that it's lucrative for. It's a great career for many, many people. I think you just have to move to the pods when you're ready to start generating ideas, and ideas that are profitable. What I tell people is that it's a mistake to move to pods too early in your career. You have to be in a place where you have a long enough track record where you know that you can make money if you're put out into the wild to do it. Before that, business development calls a lot at every point in your career, but you have to know when you're ready, and I think it really only makes sense to go over as a senior analyst.

Pari Passu: Got it. Perfect. Okay, so walk us through what went through your mind. I assume you were around 30, hopefully doing well, you have a very good seat at Citadel and a lot of upside. And then obviously you made a very big career shift. It would be great to hear what was going through your mind: what you were thinking?

Later we can talk about the specific company you founded, but maybe we can start with the inklings of entrepreneurship and how you thought about the choice, because obviously you were in a great seat and basically impossible to get back once you leave.

Mohit: I never dreamt of entrepreneurship, it was never something I really wanted to do. It was really just recognizing how big the AI opportunity was, and realizing how transformational AI was going to be for everything around us, that led me to say, all right, I have to be involved in this. It was recognizing, really, at the end of the day, that this is going to be huge, and there are going to be ways to be impactful and build something meaningful and long-lasting in this industry. And I wanted to give it a try. Even just for personal development reasons, to see if this is something that I would be good at, and to take a risk. I've been taking risks my entire career. And so this was another opportunity to do that.

Pari Passu: Awesome, that's very inspiring to hear. So you realized AI is going to be a massive force. But you are a career hedge fund analyst. What do you do next?

Mohit: The first thing that you do is you go talk to as many people as possible. If you're thinking about quitting your lucrative hedge fund job, the first thing that you need is perspective. And you need to get as many perspectives as possible from people that have done it before, from people that are doing it right now, from the people that you hope to sell to, from the people that you hope to raise money from. You need to have as many of those conversations as possible before you make the jump. So that's really what I did next. I spent a few months, part of my garden leave, really just trying to talk to as many people as possible, both from the perspective of, am I doing the right thing, before I left, and then later, all right, if I'm stepping out to build a business that solves a problem for somebody, what are those problems, and how can we most effectively solve them with technology?

Pari Passu: Got it, that makes sense. So obviously you went on and founded Constellation. Can you give a little bit of background about how Constellation came to be, and maybe start with the evolution that I've seen since the founding?

Mohit: So when we started Constellation, we went out, we spoke to as many people as possible, and we were trying to figure out what the right problem was to solve in credit. And the problem that we landed on was credit document analysis. Document analysis is expensive, and as an industry, we spend billions of dollars a year writing, reading, and ultimately trying to circumvent credit documentation. And we thought this was something that AI was going to do really well. And so we went out, we built a system that helps analysts and lawyers read and understand credit documents faster and more accurately than existing solutions. And we launched it publicly in January. We got a lot of really good traction. We went out, and we raised a little over $3mm. Today, we are a team of five. We have a forward deployed engineer, one backend engineer, one AI researcher, our CTO, Carter, and myself.

And then we started selling this to people and started building it in public. Very quickly, a lot of things changed. One of the things that changed was that horizontal AI got much better, and people were willing to wait for horizontal AI to continue improving instead of buying purpose-built software to solve individual problems, so-called point solutions. But I think more importantly, as we talked to a lot of people in the industry, we realized that the scope of the problems was just much bigger than credit document analysis. And there were other problems out there that were really worth solving. 

I believe AI employees are coming sooner rather than later. We can see the pace at which things are developing, and it's very clear that we are going to have fully autonomous knowledge workers in the next 12 to 18 months. I think that's a conservative estimate at this point, given how quickly everything is moving.

Pari Passu: Tell me more. What does it mean to have an AI employee?

Mohit: Think about how you work with somebody who's fully remote today. You email that person, you Teams or Slack that person, you pick up the phone and you call that person, you invite them to a Zoom call. But those are all things today that AI can do. And once the right people put it together in the right format, and I think there will be multiple companies that win at this, it won't just be one, for a lot of different reasons, that is going to be the first fully plug-and-play employee that can do everything that a regular person can, except for coming to happy hour. And I think that's the direction, particularly in a lot of administrative jobs or back-office jobs, that we're headed.

Pari Passu: Let’s take a step back, especially for people who are not deep in AI, what are some of the other big things that you realized that created a barrier or would need help from specialists to implement?

Mohit: Despite extraordinary advances in artificial intelligence, putting AI to work remains surprisingly difficult. People are assembling models, tools, integrations, prompts, skills, and workflows into something useful. Then they have to supervise these systems and explain how to perform tasks that a competent junior employee could learn through observation. We think this is fundamentally the wrong abstraction. When a firm hires an analyst, it doesn't write a program specifying every action that employee should take. It provides context, demonstrates good work, delegates responsibility, and offers feedback and AI should work the same way.

I mean, go back to personal computing in the 1970s. In 1975, owning a computer often meant purchasing a kit like the Altair 8800, assembling hardware, and learning to operate a machine built primarily for enthusiasts. The Apple II, introduced in 1977, helped change that. Computing became a more complete, accessible product rather than a technical project. We think AI is about to go through the same thing. Building a useful AI agent today resembles assembling an early personal computer. Prompt engineering and manually authored skills still require users to describe, in non-technical terms, how the machine should behave, but the next generation is going to eliminate much of that burden and deliver an experience that feels like working with a junior that’s always getting better.

Pari Passu: Is that what you are working on now?

Mohit: [Laughs] I won’t give away too much. But if you’re a hedge fund, private equity firm, independent sponsor, search fund, or just an investor and you’re bottlenecked trying to look at the universe of opportunities despite leveraging Claude, we should be in touch.

Pari Passu: While I can see why building AI systems is very valuable, PMs and all professionals are very busy, so it is not always easy to spend time and money on this. Why does this matter and why should people care? 

Mohit: Let me explain it this way. For anyone who's in a buy-side or capital allocation seat, they'll appreciate this, and then I'll extend the analogy further. There's a concept of sources of alpha, and there's something called BAIT. It's behavioral, analytical, informational, and technical. 

Informational advantage has gone away over the last 20 years because we've gotten more tools like 9fin or Daloopa or Cognitive Credit. That informational advantage has gone down over time, where now on most public names, you're not going to have better information than the other person. Even most sources of alternative data are pretty well understood. If you think back to 2010, 2011, when we were at the peak of alternative data, people were looking at satellite photos of cars in parking lots. That alpha got competed away very quickly. So informational alpha is very small. So recently, what we all relied on was analytical alpha.

Analytical alpha was the idea that, given the same set of information, we would reach different conclusions, because some of us were better or worse at analyzing that information or had better or worse frameworks for analyzing that information. But as frontier intelligence becomes a commodity, that analytical alpha is going to go to zero. The frontier models are just as capable as you are in most things. And with the right context and the right harnesses, they can get to many of the same conclusions, or at least help people get to those conclusions faster, or give them more time to spend on the detail of analysis that really makes the difference. So analytical advantage is also going to get competed away as frontier intelligence gets smarter and smarter. And so then the question is, all right, how are firms going to compete with each other? 

You're really left with behavioral: are you able to hold positions longer, are you able to be disciplined in your investing strategy? Or technical, and in this case technical means things related to market structure: are you large and can you write large checks, or can you play in certain parts of the private credit sphere that other people can't, etc.. 

And so eventually, within those same buckets, all the mega-cap firms, all the single managers, all the multi-managers, they're all ultimately just going to collapse to each other. There could be very little differentiation if informational and analytical advantages are competed away. And so why should people be focused on AI right now? Because everybody else is. And because everybody else is focused on AI, and because that advantage is getting competed away, people are getting faster and they're getting smarter, and the people that don't are going to be left behind. 

The people that are not deploying AI in their organizations are the equivalent of people that are saying, “You know what, it's okay, we don't need to get access to SEC EDGAR financials, we can just download the annual reports, or ask the company to mail us annual reports.” Technology is moving, and no one would deny that access to Bamsec or Daloopa or Cognitive Credit makes their job better, and their job would be really hard if they didn't have those, and they would be at a competitive disadvantage if they didn't have those tools. That's the same direction that we're going with AI.

More importantly, you’re making my point for me a little bit: it’s time-consuming to set up all of these AI systems. It’s not easy. Everyone has a day job already. So systems that you can drop in and learn from you in real time become even more valuable in these situations.

Pari Passu: This is really fascinating. I have two more questions I want to ask you about. The first one is, obviously, you interact a lot more with AI than an investor, a banker, or a lawyer does. What are some other things you've realized and would like to share? 

Mohit: When you deal with AI at the frontier, you realize how powerful it really is. If you spend any amount of time talking to researchers or talking to the security community, you'll realize how much of a real threat AI represents. But also why it's good for companies like CrowdStrike that do cybersecurity, and why those companies are really, really important in today's climate. So that's the thing that I was surprised by as I spent more time with AI.

Pari Passu: I also think you have a very unique overlap, in that you're both a credit investor and now very competent in AI. For all the credit analysts looking at the BSL software universe, any advice for them on how to navigate the existential terminal-value worries for all these software names?

Mohit: Look, I think some pieces of software will actually have existential problems at the end of the day, but the reality is, I think it's called Jevons paradox. AI is going to create a lot more opportunity and create a lot more economic value than it will destroy. The question is just: is the piece of software or the company that you're looking at adapting well to AI, and how does their technology fit into an AI-first infrastructure or architecture? So things that are systems of record are very, very durable. Things that are based on friction costs, things that you just don't cancel, are very undurable. Those are going to go away. Things that are infrastructure, that connect the pipes or are the pipes themselves, are extremely durable. So you need to think about it from the perspective of, if I were restarting a business tomorrow, and I would want it to be as AI-native as possible, if I was building an AI-native consultancy, or banker, or AI-native accounting firm, or whatever it is, would I need this piece of software? That's really the question that people need to be asking. 

I think what people forget is that building things in Claude Code is very fun. That's a big part of why people are doing it internally, because there's a really satisfying feedback loop to it. That's definitely why I took a lot of engineering classes in college, because it's very fun. But the counter to that is that it's not very fun to maintain software. One of the things that we run into a lot is firms that have built their own internal systems, systems that are so complicated that they could rival things built by Oracle or Palantir themselves, and then they say, well, we don't know how to maintain these systems, because we all have day jobs, but we still need to maintain the system that we've now started to rely on. 

And so part of why people buy software is because sometimes it's easier to buy software than to build and maintain software. And I think we've just now gotten to the second side of that yet. If you're a credit analyst and you've built this Claude skill for how to read credit documents, that's very good, and you're going to get a lot of use out of it. But the reality is when other people start using it, and it starts breaking, and people say, well, can you change this, can you do that, why is this not working, then suddenly you're like, all right, I've got two jobs instead of one.

Pari Passu: This was very fun. Let me ask a bit of a different question to end. A finance career is often seen as prestigious, it's very stable (at least compared to entrepreneurship), while still having a lot of upside. Has anything surprised you from making the switch to entrepreneurship, ideally both on the positive and the negative? What are some of the things that you wish you knew earlier? 

Mohit: Something that surprised me in a positive is that if you have a growth mindset, you will figure out how to do difficult things. This was not something I set out to do, but I figured it out along the way, and that's the power of having a growth mindset. What was surprising to the negative is that running a startup is like chewing glass every single day and then smiling at the of it. That's really how it feels, most days. So that was a surprise, but maybe it should have been expected. My advice to people is that you should try this once in your life. Life is long, there are lots of opportunities. But when you're young, it's the time to take swings, and there's really no time like the present, when so much economic value is going to be created. So even if you don't do something yourself, if you go join somebody else who's doing it, or even just ideate on what the art of the possible is, that's going to be incredibly valuable to you, and incredibly rewarding.

Pari Passu: Beautiful ending. Thank you, Mohit. I learned a lot, but it was also very fun for me to do. So thank you for taking the time.

Mohit: Of course. Great to do this.

Disclaimer: Pari Passu is compensated by Constellation for distribution and marketing. 9fin and Cognitive Credit are active Pari Passu sponsors. Daloopa and Hebbia are former Pari Passu sponsors. These relationships had no bearing on the accuracy or substance of this discussion.