Michael Orabueze

ex-Balyasny commodities quant. Building AI for hedge fund operations. SF

[Blockhouse Archives @BlockhouseClips]

Balyasny's Global Head of Quantitative Research Giuseppe Paleologo (@__paleologo), formerly of Citadel, Millennium and HRT, on how the three firms actually differ:

"I'll actually quote a co-head of one of these firms, without mentioning his name, who once told me, 'Citadel is Singapore. It's centrally managed, it's very efficiently managed. Everything is clean and working.'"

"'Millennium is the United States because it's federated. If you want, you could say it's siloed, it's messy, it's big, but things work out in both models.'"

"I would add that HRT is a little bit like Wakanda. So HRT is smaller, in a sense it's cooler. The human capital, the quality of people, I think is probably the highest among the firms that I've worked at."

[Michael Orabueze @m_orabueze]

@hkuppy Survived Monday, but Thursday was the real test for ops: margin calls and a stack of PB and admin breaks the next morning.

[Michael Orabueze @m_orabueze]

@jam_croissant @NoelSmith @volatility911 Two launches at once means new ISDAs, PB and admin onboarding, and OTC confirms flowing from day one. Curious who's building the ops side.

[Brett Caughran @FundamentEdge]

As a PM, onboarding a new, fresh junior analyst is a terrible experience. At best, it's a six-month distraction, and that junior analyst is almost always a near term cost center to the business. Workflow by workflow, you have to show that junior analyst what to do, and very closely monitor that analyst's work for errors, providing direct feedback along the way.

And that’s the easy part. The hard part is enforcing how to think. Un-training the overconfidence out of that junior analyst, and pushing back harshly when the analyst makes a claim unfounded by the primary research.

If you asked a PM what the ROI of training a junior analyst is around month four or five, without looking ahead, they would say it's terrible!! I hate this, and I never want to bring on a new junior analyst again!

Usually around month six, that analyst can then be trusted to conduct workflows on his or her own. Okay, this is a little bit better. The work product is good now. I can loosen up my oversight a bit.

And the ROI gets better.

Usually after two years, good analysts cross a threshold. And they start to become standalone insight generators, and talented stock pickers (many never make this move as some of it is innate)..That two years of feeding into that junior analyst starts to pay off, and at the end of the year, you can start to notice that analyst's impact in the P&L of the business. At this point, you're grateful for the two years you spent training that analyst, and while the junior analyst was more expendable to the business, the talented senior analyst is incredibly critical to your efforts.

I think that's where we're at now with agents.

The ROI of almost all motion on AI for investment research over the last three years has been definitively terrible for public market investors. I don't think it has really bent the alpha curve anywhere, consistently and at scale (certainly hasn't shown up in P&L at scale anywhere). And I've heard many complaints about all the hours wasted for no payoff. But also, we shouldn't expect a junior analyst to bend the alpha curve.

But a senior analyst? A well-trained, proactive, end-to-end system. That's much more interesting. That is an entity that can start to bend the alpha curve in a portfolio.

The most forward-leaning public market investors are very close to having swarms of senior analyst talent at their fingertips. My belief is that over the next 6 to 18 months, this will start to bend the alpha curve in a noticeable way.

[Michael Orabueze @m_orabueze]

big down days aren't just a PnL story. margin calls jump, collateral disputes pile up, and the PB and admin files don't match your book the next morning. ops pays for volatility a day later

[InfraMarginal @infra_marginal]

Millennium has hired Max Iakovlev as a senior energy portfolio manager

Iakovlev left Balyasny in Q1 after six years. Bloomberg reported in March that he and Toby Sheppard had left Balyasny's London power and gas team.

Before Balyasny he spent five years as one of Mercuria's senior derivatives traders.

He started at Goldman Sachs, where he spent more than eight years and rose to managing director.

[Hedge Fund Clips @HedgeFundClips]

Ex-Point72 Head of Prop Research Kirk McKeown on what the real edge was at SAC and Point72:

"When you go to a place like Point72, which was SAC when I got there, it was about understanding when stories changed one degree instead of 10."

"Understanding when those stories were changing was the competitive advantage, I think, of that firm."

"Where is the market, where is the world, and where is the story relative to that? And is it still in line with where the story was, or has it changed based on new information that you get, or a change in tone with the management team?"

"So it's really around capturing the nuance. It's about having a scaled organizational framework where you have a repeatable process that ends up being the thing that creates the alpha."

[The Quant Clips @TheQuantClips]

The human threshold:
Bridgewater's managing CIO Greg Jensen

"If you take the period where AI was going from the intelligence of a squirrel to the intelligence of a monkey, it's actually a huge deal on a scientific basis and has no economic impact."

"But all of a sudden, if you get intelligence above human capability, the implications are massive."

"For a long time, chess machines couldn't beat the best humans. Now they beat the best humans and no human will ever beat a good chess machine ever again."

"For the most part, all the major math breakthroughs in the future will be done with AI. Humans are mostly slowing you down at this point."

"No cyberwar, no cyberdefense will occur that's not AI driven."

"Largely all of these domains are changed forever and they're never going back."

[Matt K @KmateoK]

Looking to add another analyst at a growing SF Bay Area HF.

Global generalist fundamental l/s equity strategy + proprietary sourcing and data science platform.

If you have a passion for public markets, reach out. 3-10 years experience req’d. DMs open, retweets appreciated.

[The Quant Clips @TheQuantClips]

15x leverage and a 2% knockout can turn commodity stress into a rush for the exit.

CIBC global head of QIS structuring Kimberly Gallant explains:

"I think there's a difference between something going from alpha to a genuine risk transfer and having a fair price of risk and something being crowded, right?"

"Then it's gone through most of the last 20 years where it's behaved as a reasonable factor delivering a Sharpe ratio of about one, which by the way, is still pretty great for something that is broadly uncorrelated with, uh, equity risk premia, bond risk premia, and other factors."

"We started to see that contagion spread to other areas, and that's a sign of crowding"

"some other banks had sold fifteen times levered structured notes linked to the commodity curve pl- complex with a knockout that was down, uh, if the market, if the strategy was down two percent."

"And so that is exactly the type of mechanism that leads to not only has this thing become crowded because you've made it very easily accessible to the market with leverage, but you will also add a, a trigger point that a real insurance event will cause a rush to the exit."

[The Quant Clips @TheQuantClips]

Why fundamental analysts don't day-trade:
Point72's founder Steve Cohen

"I think that there are plenty of quants that trade second by second, okay?"

"Most of my fundamental discretionary equity, the guys that do the analytical work don't trade minute by minute."

"I mean, your transaction costs would kill you when getting in and getting out of the market is expensive."

"And markets have become more and more efficient, although there's a rise of traders who buy out-of-the-money call options as soon as they hear AI."

[Benn Eifert 🥷🏴‍☠️ @bennpeifert]

If you have been a derivatives trader or salesperson and are interested in a client facing role please reach out!

[Hedge Fund Clips @HedgeFundClips]

Ex-Citadel quant Renee Yao, founder of Neo Ivy Capital, on why her AI models would never use wet grass to predict rain:

"You should let the model identify the features themselves, instead of you, as a human researcher, telling them which features you're going to use."

"Traditional machine learning also uses features, and the features they use are mostly linear in nature and mostly only associative in nature."

"Your traditional model will tell you, hey, this is the perfect feature. Let's just use grass getting wet to predict the rain. But we know it's not causative."

"It actually won't use that. So we're only looking at nonlinear ones."

[Michael Orabueze @m_orabueze]

Most ops tech is built for the sunny day.

The question that separates good systems: when one trade fails, does it block the queue, or get quarantined with a full audit trail while everything else flows?

[Hedge Fund Clips @HedgeFundClips]

Balyasny's founder Dmitry Balyasny on going from a 'printer reader' to AI:

"When I started trading at Schoenfeld, we had a person whose job it was to be the printer reader. And so you didn't get your fills electronically, you would get your fills coming back on a printer. And some poor guy's job was to read out your prints."

"You contrast that with all the AI and data technology today that we and others use for trading and investing, and it's just a tremendously different world."

"Although I think AI will likely be a larger, more substantive change over time, we're coming from a place where everybody's already enmeshed in technology in so many different ways."

"So I think it'll be more profound over the long term, but the change feels a little bit less than it did at the time."

[Hedge Fund Clips @HedgeFundClips]

Balyasny's founder Dmitry Balyasny on why today's clunky AI looks like the early internet:

"AI is good now, it's clunky still."

"Remember trying to download video on the internet in the 2000s? It didn't work very well. And a lot of people wrote it off."

"You look years later and Netflix didn't write it off, and they built a gigantic company, and the technology got better and better."

"So I think AI is on a faster curve, so it's just going to be improving faster, faster, faster."

"Hard to say who's going to make money and which companies are going to do well, that's a whole separate question. But the technology and the usefulness, I'm convinced, is going to be great."

[Ted Seides @tseides]

"Every hour a GP spends managing a CV is an hour they're not spending managing my commingled fund."

— Scott Ramsower, Head of Private Equity Funds at Texas Teachers

https://t.co/lFK45hB2Sm

With thanks to @AlphaSenseInc, @Juniper_Square, & Ridgeline.

[Christina Qi @christinaqi]

Databento is a 100% remote company, which in finance is like... insane. During new hire orientation, I do everything I can to keep us 100% remote by setting expectations:

Remote does NOT mean "nomadic." It doesn't mean traveling, cameras off, wondering where you are. You're either 100% on vacation or 100% working. We expect a stable work schedule. "Don't be that person who breaks the trust."

Remote DOES mean the privilege of flexibility in your daily routine. You can walk the dog, pick up the kids from school, go to the gym before rush hour. So long as you communicate w/your team and make your meetings on time.

A few more quotes:
"Don't be that person who tests company boundaries." (Especially during this 'anti-work' era, like sure do the bare minimum for corporate overlords at Amazon or McDonalds, but I promise not every company is evil. 🥹)

"If a company has to change a policy because of 1 person, don't change the policy. Fire the person." (I hope folks can relate to this.)

Ideas flow better in-person, so we have to do offsites to compensate.

Also not every type of company is suitable for remote work (e.g. labs & warehouses have to be irl).

Not sure if I covered all my bases, but let me know if there's any questions. My previous company was 100% in-person so I've been on both sides of the table.

[CAMMS Capital @CAMMSCapital]

CAMMS Capital is live. Backed by @Braiins.

We spent five years trading commodities inside ETG. Now we're launching as an independent company, CFTC-registered and NFA member.

Built inside ETG, a global commodity conglomerate with expertise across multiple industries, we developed and traded a multi-strategy commodity platform before spinning out. Our research partnership with ETG continues, giving us a direct window into supply chains, trade flows, and fundamental market drivers.

Today, we're launching as CAMMS Capital, acquired by Braiins and rebranded from CAM Multi-Strat Asset Management LLC.

We trade across agriculture, energy, metals, freight, and macro through three approaches: discretionary (driven by physical market intelligence and ETG research), hybrid (quantitative and fundamental, regime-adaptive), and systematic (fully rules-based).

The founding team remains in place, incentives aligned:

→ @StephaneCAMMS, CEO. 32 years at Louis Dreyfus, InVivo, ETG.

→ Salim Boutaleb, CIO and Head of Quantitative Trading. 15 years in systematic commodity strategies.

→ Rafael Molinero, COO. 28 years in the hedge fund industry, Chicago Booth MBA.

75+ years of combined commodity experience.

Building diversified commodity exposure means sourcing multiple specialist traders and running a risk framework across them. We do that work, drawing on decades of relationships and market access, so one allocation gives you exposure across 30+ commodity markets.

Institutional infrastructure from day one. KPMG audit, Northern Trust custody, SS&C administration, StoneX and Marex as prime brokers.

Commodity Intelligence, Amplified.

Welcome to CAMMS Capital.

[George Hadjia @GHadjia]

MasterBrand $MBC just announced a cheeky immediate CFO transition the week before their earnings print. Never a good sign. We wrote up this short in our latest quarterly letter. See below.