How To SpeedRun a Quant Trader Career With Ai

*Inside Citadel, a backtest now runs on 10,000 cores in 5 minutes instead of overnight on 100 servers. At Man Group, LLM agents generate 50-100 trading ideas a day. At Bridgewater, a 17-person team is building a system that "works like millions of 80th-percentile associates in parallel." The profession is changing faster than at any point in its history. And for the first time, the path into it is open beyond top universities.***

*A speedrun in gaming means completing a game in the minimum possible time by using every legitimate optimization the system allows. Not cheats. Not shortcuts that break the experience. Just deep knowledge of the system plus tools that turn dead time into progress.*

This article applies the same frame to quant trader. The "game" is a hard, traditionally gated profession. The "skips" are AI tools that compress months of solo grinding into days of guided learning. Some routes still require the full classical foundation. Others have been shortcut by the existence of LLMs and AI infrastructure that didn't exist three years ago.

The lens: not "how to get rich in quant trading," but the optimal AI-augmented path through a hard technical profession in 2026.

The numbers you need to see first

*What the industry pays in 2026:*

• Entry-level total comp at Citadel: $336k-$642k including bonus
• Jane Street, average per-employee payout in 2024: $1.4M
• Five Rings, base for quant trader: $300k
• IMC Trading interns: annualized equivalent of $240k+
• Five-year benchmark at top prop shops: $800k-$1.2M per year
• Top traders with strong P&L: eight figures per year

*What's happening in the quant industry right now:*

• Citadel Securities: migration to Google Cloud, access to 1M+ cores for parallel backtesting
• Man Group ($151B AUM): launched AlphaGPT - multi-agent system replacing a research pod
• Bridgewater: AIA Labs (17 people) already trading client money via LLM workflows
• Point72: GPT in a locked Azure V-Net, follow-the-sun hubs in Warsaw and Bengaluru
• D.E. Shaw: LLM Gateway brokering requests to 24+ external models with PII stripping

*Speedrun time (with AI):*

• With STEM background, no PhD: 18-24 months of focused work
• From scratch: 3+ years (through MFE - the standard path)

*Two connected shifts make the speedrun possible: the industry is moving to AI, and the same AI has opened access to the profession for people without an MIT diploma. Both are happening right now.*

What a quant trader actually is, and why this role specifically

*There are 4 different roles in the quant industry that often get confused:*

*This article is about the trader role, not by accident. It's the most accessible for a wide audience:*

• A STEM bachelor's is enough. Jane Street explicitly states: "prior knowledge of finance or economics is not expected or required." Two-thirds of their intern class is CS and math, not finance.
• No PhD gating - unlike research positions at Renaissance/Two Sigma/D.E. Shaw
• Highest variance career path. Exceptional traders make 8 figures; ordinary ones make $200k-$400k entry-level and $800k-$1.2M after 5 years
• The most "real-time mathematical" role - probability puzzles, mental math, decision-making under pressure

*What a trader does day to day:*

• Market making: quotes bid/ask, manages inventory, hedges risk
• Implements strategies built by researchers
• Makes fast decisions: pricing nuances, order book dynamics, rare events
• Manages P&L on their book in real time
• Pairs with researchers to improve models from live data

*The core frame: a trader doesn't predict the market. A trader finds situations where the mathematical probability is shifted in their favor, sizes the position correctly, and repeats the process thousands of times until EV accumulates into real returns.*

*It's a casino from the other side of the table. The casino doesn't predict any single bet's outcome. It runs the game with a small mathematical edge and lets the law of large numbers do the rest. Quant trading works the same way*.

How AI is changing the work of the quant industry (the part roadmaps usually skip)

*Most "how to become a quant" content skirts the most important thing: the 2026 profession is no longer the 2020 profession. If you're learning by a 2018 plan, you're learning for a role that has already transformed.*

Here's what's actually happening at top funds. Sources: Business Insider investigations, Resonanz Capital research, industry conferences.

AI as a research co-pilot

*This is the baseline level, already in place at all top firms. Every research analyst has an internal LLM tool:*

• D.E. Shaw - DocLab. Vector database over all internal research notes. A query like "what did we say about energy markets in the last 2 years?" returns a structured answer with confidence scores and audit hashes
• Point72 - GPT in locked Azure V-Net. Access to powerful models without leaking data outside
• Balyasny - in-house gateway. 10 data sources through a unified AI interface
• Bridgewater - LLM co-pilots for compliance and research. 70% reduction in manual review time

*This is the level you can enter today. Use Claude or ChatGPT for research work - and you have the same capability as a junior at a fund, minus the private data.*

AI generates trading ideas

*Man Group's AlphaGPT(https://www.man.com/insights/what-ai-can-do-for-alpha) is the brightest example. The system mimics a human research pod but runs at machine speed:*

• Agent 1 (idea generation). GPT-4 fine-tuned on historical research notes. Produces 50-100 strategy ideas per day, combining current market observations ("gold/oil ratio is widening") with pattern recognition over history
• Agent 2 (implementation). Converts a natural-language idea into Python code. "Test momentum strategy on energy stocks using RSI and volume confirmation" -> ready-to-run backtest script with data loading, signal generation, performance metrics
• Agent 3 (risk evaluation). Analyzes backtest results for overfitting, factor exposures, robustness

*What this means for a Man Group trader: the idea-to-P&L cycle compresses from weeks to hours. Previously, an idea went: human researcher -> code via quant developer -> review through risk -> submission to PM. Now: human formulates -> AI implements -> human verifies.*

Production-grade infrastructure for AI trading

*Citadel Securities migrated quant research from on-prem to Google Cloud. Numbers from their public decks:*

• A backtest that took 8 hours on 100 dedicated servers
• Now runs in 5 minutes on 10,000 ephemeral cores
• Wall-clock speedup: 96x
• Team tests 50+ strategy variations per day instead of 3-5

*Edge through infrastructure: 10x more hypotheses tested per unit of time.*

Full replication of the human process

*Bridgewater AIA Labs(https://www.bridgewater.com/research-and-insights) under Greg Jensen is the most radical experiment. Goal: fully replicate Ray Dalio's macro-investment process through machines. Running on AWS EKS, already trading client money.*

Insiders describe the system as "working like millions of 80th-percentile associates in parallel." This is the first public demonstration that a macro fund is ready to entrust the entire investment loop to an LLM-heavy workflow.

NLP and alternative data

*Point72 deploys NLP models for analyzing earnings calls, regulatory filings, news sentiment. This is no longer a competitive edge - it's baseline expectation for any major multi-strat fund.*

What this means concretely: a 2026 trader doesn't sit reading 10-Ks. The trader queries an AI system that has read every 10-K from every year and highlights changes in wording between quarterly reports. Signal in those wording shifts is real alpha being actively exploited right now.

What to take from this (critical for the roadmap)

Previously, quant trader skills were math, mental math, market intuition. That's still the foundation. But in 2026 there's now AI fluency, which has shifted from "nice to have" to baseline expectation.

*What should be in your skill set, beyond classical foundations:*

• LLM-driven research: how to structure prompts, how to validate outputs, where AI gets it wrong on financial data
• Vector databases and RAG: understanding how systems like D.E. Shaw DocLab work
• Agent workflows: how multi-agent systems like AlphaGPT are built
• Fine-tuning and prompt engineering for finance: specifics of financial data and text
• AI risk and model validation: how to verify that AI hasn't overfit or hallucinated

*This used to be exotic. It's now the reality of interviews at Citadel, Man Group, Bridgewater, Point72.*

The ladder: 7 steps from zero to offer

*The biggest mistake most people make is trying a vertical jump. Submit a resume to Citadel without credentials, get rejected, conclude the path is closed. The path isn't closed. It's a ladder where you can't skip steps.*

*Each step is a prerequisite for the next. Step 5 is new in 2026 - it didn't exist in 2020 roadmaps.*

Each step in detail below.

Math foundation

*The longest and most unforgiving step. At an interview at Jane Street or Optiver, it shows in 5 minutes whether you went through this honestly or jumped past it.*

*5 layers in strict order:*

Probability (8-12 weeks)

The most important layer. Everything in quant trading reduces to one question: what are the odds, and are they in my favor?

The key concept here is conditional thinking:

For example: the unconditional probability that the market moves sharply after an economic announcement is 40%. Conditional on options implied vol being elevated beforehand, the probability of a sharp move becomes 68%. Those 28 percentage points are real alpha.

*Bayes Theorem - how to update beliefs as new information arrives:*

Traders who update beliefs faster and more accurately consistently outperform those who don't.

*Expected value and variance - two numbers you'll think about for the rest of your career. Positive EV + survivable variance = money. Position sizing too large relative to variance = bankruptcy before EV gets a chance to work.*

*Resources:*

• Blitzstein & Hwang - Introduction to Probability - free from Harvard(https://stat110.hsites.harvard.edu/), chapters 1-6
• Harvard Stat 110(https://projects.iq.harvard.edu/stat110) (Joe Blitzstein) - best probability course in the world, free on YouTube(https://www.youtube.com/playlist?list=PL2SOU6wwxB0uwwH80KTQ6ht66KWxbzTIo)
• DeGroot & Schervish - Probability and Statistics (the standard)

Statistics (4-6 weeks)

Probability is theory. Statistics is the practice of "listening to data." The main lesson: most things that look like signal are actually noise.

Hypothesis testing: you've built a strategy, the backtest shows 15% annual returns. Real edge or luck? If you tested 1,000 random strategies, 50 of them will show a "significant" result at the 5% level by pure chance. This is the multiple comparisons problem - the main reason backtests look great and live results are terrible.

Linear regression: regress strategy returns on known risk factors, look at alpha (intercept). If alpha == 0 after accounting for all factors, your "edge" is just hidden exposure to something already known.

*Resource:*

Wasserman - All of Statistics (fast and dense, chapters 1-13).

Linear algebra (4-6 weeks)

*Portfolio construction, PCA, neural networks, covariance matrices. The main formula:*

where `w` is the weight vector and `Σ` is the covariance matrix. The heart of Markowitz portfolio theory.

Eigenvalues show what actually matters in the covariance matrix. In 500 stocks, the first 5 eigenvectors explain ~70% of all variance. The rest is noise. This is the foundation of factor investing.

*Resource:*

Gilbert Strang - MIT 18.06(https://ocw.mit.edu/courses/18-06-linear-algebra-spring-2010/) (free on MIT OCW) + his book Introduction to Linear Algebra. For intuition - 3Blue1Brown - Essence of Linear Algebra(https://www.3blue1brown.com/topics/linear-algebra) (free on YouTube).

Calculus and optimization (4-5 weeks)

Almost every problem in quant finance reduces to maximizing something subject to constraints. The key here is convex optimization: a convex problem has a unique global solution that can be found efficiently.

*Resource: *

Boyd & Vandenberghe - Convex Optimization - free from Stanford(https://web.stanford.edu/~boyd/cvxbook/), chapters 1-5. The author's course - Stanford EE364A(https://web.stanford.edu/class/ee364a/).

Stochastic calculus (6-8 weeks)

After this layer you can derive how financial instruments are priced from first principles.

*The central idea: in a world with randomness, the square of a small random increment is not negligible the way it is in ordinary calculus. This produces Ito's Lemma - the chain rule of stochastic calculus. Apply it to an option price and you get the Black-Scholes equation:*

What's remarkable here: the expected return of the stock disappears completely. The option price doesn't depend on where you think the stock is going. Only on how much it moves.

Resource:

Steven Shreve - Stochastic Calculus for Finance II(https://link.springer.com/book/10.1007/978-0-387-22527-2). The industry standard.

How AI changes all this learning

Before: stuck on a proof in Shreve - you spend days alone, read forums, maybe find an explanation. Often you give up.

*Now: a 30-second prompt:*

You get an explanation in 5 minutes. The same applies to probability paradoxes, regression derivations, intuition behind eigenvectors.

Programming

In parallel with math. Not sequentially.

In the quant industry there are two different programming worlds, often confused:

*Research programming - Python, clean code, statistics, ML. The trader's and researcher's main tool.*

*Production programming - C++ or Rust, microsecond latency. For quant developers; for traders, basic understanding is a plus.*

Python for traders

*What to master:*

• NumPy(https://numpy.org/) - vectorization (10-100x faster than ordinary Python)
• Pandas(https://pandas.pydata.org/) - time series, multi-index, groupby
• Polars(https://pola.rs/) - modern Pandas replacement, 10-50x faster on large datasets. In 2026 - must have
• SciPy(https://scipy.org/) - statistics, optimization
• statsmodels(https://www.statsmodels.org/) - statistical tests
• cvxpy(https://www.cvxpy.org/) - convex optimization
• xgboost(https://xgboost.readthedocs.io/), lightgbm - tabular ML

*Resource: *

Wes McKinney - Python for Data Analysis(https://wesmckinney.com/book/) (from the author of Pandas, free online).

SQL and big data

*What to learn: joins, window functions, CTEs, parquet handling, basic understanding of ClickHouse(https://clickhouse.com/) (the standard for financial time-series).*

Resources: Mode Analytics SQL Tutorial, LeetCode SQL.

C++/Rust

*For a trader, not critical, but a plus. C++ has been the HFT standard for decades. Rust is gaining momentum in 2026 - NautilusTrader(https://nautilustrader.io/) (open-source production-grade trading platform) uses a Rust core + Python API; this pattern is becoming the standard.*

How AI changes this learning

*Programming is the most AI-transformed area in the entire roadmap:*

1. Code review. Previously you needed a senior developer. Now: "here's my backtester, find look-ahead bias, find survivorship bias, find where transaction costs are wrong" - answers in a minute, more accurately than half the seniors

1. Debugging. Ask for "5 hypotheses for why this test fails" instead of "fix the bug." Forces you to think, accelerates you massively

1. Reproducing algorithms. Take an academic paper, ask for pseudocode, implement yourself. Used to take weeks. Now an evening

1. Boilerplate. Data loading, visualization, configs - written in 30 seconds via prompt. Frees you to focus on core logic

Finance and market structure

STEM candidates are often weakest exactly here. They arrive with strong math but don't understand option Greeks or yield curves.

Foundations

• Hull - Options, Futures, and Other Derivatives(https://www.pearson.com/en-us/subject-catalog/p/options-futures-and-other-derivatives/P200000005938). The main book on derivatives. Everyone reads it
• Bodie, Kane, Marcus - Investments. Standard textbook
• Time value of money, NPV, bond pricing, duration, MPT (Markowitz), CAPM, Fama-French

Options and volatility

*A trader-specific area:*

• Natenberg - Option Volatility and Pricing(https://www.mhprofessional.com/option-volatility-and-pricing-advanced-trading-strategies-and-techniques-2nd-edition-9780071818773-usa). Specifically for traders
• Greeks (delta, gamma, vega, theta, rho) - interpretation and use
• Implied volatility, vol surface, vol smile/skew
• Put-call parity, exotic options conceptually

Market microstructure (critical for traders)

• Cartea, Jaimungal, Penalva - Algorithmic and High-Frequency Trading(https://www.cambridge.org/core/books/algorithmic-and-highfrequency-trading/1CB57C5ED7C2C2F36BA092A55E5B5B47)
• Order book mechanics, market makers, liquidity
• Bid-ask spread, market impact, adverse selection
• Order types and how to use them

Industry understanding

• Rishi Narang - Inside the Black Box (how quant funds are built)
• Scott Patterson - The Quants (industry documentary)
• Greg Zuckerman - The Man Who Solved the Market (about Renaissance)

How AI changes this

*Finance literature is dense and often opaque. A prompt like:*

• breaks down a chapter in an evening that previously took a week.

*Especially powerful for academic papers:*

*Previously a serious paper was a week-long task. Now an hour.*

Portfolio and competitions

*Without GitHub projects, your resume has no weight. But for trader roles there's something more important than projects: competitions.*

Trading Competitions = recruitment pipeline

*Companies watch the leaderboards. Strong performance on these brings direct recruiter DMs:*

• Jane Street Kaggle competition(https://www.kaggle.com/c/jane-street-real-time-market-data-forecasting) - $100,000 prize pool
• WorldQuant BRAIN(https://platform.worldquantbrain.com/) - pays cash for alpha signals you submit
• Citadel Datathon / Terminal(https://www.citadel.com/careers/students/datathon/) - explicitly fast-tracks winners into interview pipeline
• IMC Prosperity(https://prosperity.imc.com/) - annual trading game, top performers get interviews
• Optiver competitions(https://optiver.com/working-at-optiver/career-opportunities/) - multiple events year-round

Projects

*Minimum 3 GitHub projects with READMEs:*

*Project 1: Reproduce an academic paper from top journals (Journal of Finance, Review of Financial Studies, JFE). For example - momentum factor (Jegadeesh-Titman), any equity factor from SSRN.*

*Project 2: Statistical arbitrage strategy - pairs trading with cointegration, mean-reverting spread, backtest with realistic transaction costs.*

*Project 3: Something non-standard in 2026 - this is what differentiates you:*

• LLM-based alpha from text. Process earnings call transcripts via Claude/GPT API, extract signals, test as alpha factor. This is a hot topic in 2026 - top funds are actively building such pipelines
• AI agent for research workflow. Build an agent that takes a strategy idea in free form, generates backtest code, does sanity checks. Demo-level project few people make - and it impresses
• Reinforcement learning for market making
• Anomaly detection in high-frequency data

Where to get data

• Free: Yahoo Finance via yfinance(https://github.com/ranaroussi/yfinance), FRED(https://fred.stlouisfed.org/), SEC EDGAR(https://www.sec.gov/edgar)
• Cheap: Polygon.io(https://polygon.io/) ($200/mo - standard for serious retail), Tiingo(https://www.tiingo.com/), Alpha Vantage(https://www.alphavantage.co/)
• Institutional: WRDS(https://wrds-www.wharton.upenn.edu/), CRSP(https://www.crsp.org/), Bloomberg ($32k/year)

AI fluency for finance

*This step didn't exist in 2020 roadmaps. In 2026, without it, you don't get hired at Citadel, Man Group, Bridgewater, Point72.*

What needs to be in your skill set:

LLM for financial data

• Structuring prompts for financial analysis
• Validating AI outputs (where it hallucinates, where it's reliable)
• Understanding limitations: what AI can't do and why
• Cost optimization (when to use GPT-4, when Claude, when specialized models)

Vector databases and RAG

*Understanding how systems like D.E. Shaw DocLab work:*

• Embedding models and semantic similarity search
• Chunking strategies for financial documents
• Hybrid search (semantic + keyword)
• Confidence scoring

Agent workflows

*How multi-agent systems like Man Group AlphaGPT are built:*

• Task decomposition
• Tool use (when LLM calls Python, SQL, APIs)
• Validation loops (one agent critiques another)
• Human-in-the-loop checkpoints

AI risk and model validation

• How to check AI for overfitting on financial data
• Detection of look-ahead bias in LLM-generated code
• Model interpretability (XAI techniques for regulation)
• Black box problem and how it's handled in practice

Build something real

*The best way to demonstrate AI fluency is to build something yourself. Project idea:*

*Minimal AlphaGPT clone. 3-agent system:*

• Agent 1: generates hypotheses in free form ("mean reversion in morning hours for small-caps")
• Agent 2: writes backtest code on historical data
• Agent 3: critiques the result - overfitting, leakage, transaction costs

This can be built in a weekend with the Claude API. This is the single most impressive project you can show a Citadel recruiter in 2026.

How AI changes the learning itself

*It's recursive: you use AI to learn how to use AI. Drilling prompts:*

*After a few such sessions, you have working understanding of architectures currently being deployed at top funds.*

Interview prep

*This is its own genre. You can know everything and fail interviews. Citadel Super Days are 6 consecutive 45-minute interviews on topics from probability proofs to coding to behavioral.*

Probability puzzles and brainteasers

*The base on every interview. Main books:*

• Xinfeng Zhou - A Practical Guide to Quantitative Finance Interviews (Green Book) - the most-cited book in the industry. 200+ real problems
• Timothy Crack - Heard on the Street
• Mark Joshi - Quant Job Interview Questions and Answers(https://www.markjoshi.com/RecruitmentPage.html)

Site: brainstellar.com(https://brainstellar.com/) - interview-difficulty problems with solutions.

Mental Math (trader-specific)

For trader roles this is critical, not optional. Trainer: Zetamac(https://arithmetic.zetamac.com/).

Goal: 50+ correct answers per minute before you start applying. Without this you can't pass trader interviews at Optiver/IMC/Jane Street.

Coding interviews

LeetCode(https://leetcode.com/) (medium-hard). Especially: arrays, dynamic programming, graphs. Dynamic programming is the most common failure point in final rounds at Citadel and Jane Street.

Specifics: often asked to implement a numerical algorithm - Black-Scholes, Monte Carlo, optimization.

Behavioral - Jane Street specifics

Seriously underestimated. Jane Street deliberately makes problems too hard for one person to fully solve - they watch how you use hints, reason under pressure, take criticism.

*A candidate who narrates their thinking, considers edge cases, and continues to reason while acknowledging uncertainty consistently outperforms one who goes silent and produces a correct answer without explanation.*

How AI changes preparation

Previously, mock interviews required finding a partner, scheduling, spending an hour, and could happen at most once a week. Now: an hour every day with infinite question variation.

Drilling probability puzzles: Load up Crack/Zhou, ask to generate problem variants with different numbers. In a month you've worked through hundreds of variations.

Behavioral: The most common weak spot for technical people. "Ask me 20 typical behavioral questions at Jane Street level, evaluate my answers via STAR" - in a week turns a robot into a normal-sounding candidate.

Where to apply

*Top prop trading firms (most accessible for trader roles):*

• Jane Street(https://www.janestreet.com/join-jane-street/)
• Hudson River Trading(https://www.hudsonrivertrading.com/careers/)
• Optiver(https://optiver.com/working-at-optiver/career-opportunities/)
• IMC(https://www.imc.com/us/careers)
• Five Rings(https://fiveringsllc.com/careers)
• Jump Trading(https://www.jumptrading.com/careers/)
• Tower Research(https://www.tower-research.com/open-positions/)
• DRW(https://drw.com/work-at-drw)
• Akuna Capital(https://akunacapital.com/careers)

*Multi-strat hedge funds:*

• Citadel(https://www.citadel.com/careers/) (especially Citadel Securities(https://www.citadelsecurities.com/careers/))
• Millennium(https://www.mlp.com/careers/)
• Point72(https://careers.point72.com/)
• ExodusPoint(https://www.exoduspoint.com/careers)
• Balyasny (BAM)

*Channels:*

• Trading competitions = fastest path (see Step 4)
• Direct career sites (above)
• LinkedIn for networking (not for applying)
• University career services
• QuantNet forum(https://quantnet.com/forum/)
• Wall Street Oasis(https://www.wallstreetoasis.com/forum/hedge-fund) - process discussions
• Referrals from current employees

Main mistakes

1. Skipping math. Visible at interview in 5 minutes.

2. Ignoring mental math. Trader interviews without 50+ on Zetamac are impossible to pass.

3. Toy projects. Backtests on S&P without costs, without realistic execution. Zero resume value.

4. Not entering competitions. This is the most direct path, yet many ignore it.

5. Ignoring the AI step. In 2026, without it, top firms don't hire. Learn it in the process, not "later."

6. Ignoring soft skills. Technical people with zero communication don't pass final rounds.

7. Not applying early. "Getting fully ready" is forever. After 5-10 interviews you learn more about your gaps than after half a year of theory.

8. Using AI wrong. People either don't use it (losing 5-10x acceleration) or over-rely on it (never learning to think). AI accelerates learning, but you still have to do the thinking.

Readiness checklist

If under 80% checked - too early to apply.

*Math:*

• I solve medium probability puzzles in 5-10 minutes
• I can derive Black-Scholes from stochastic calculus
• I know time series methods (ARIMA, GARCH)

*Programming:*

• Clean Python with typing, NumPy/Polars vectorization
• Pandas advanced (multi-index, time-aware operations)
• SQL on large tables

*Finance and microstructure:*

• I can explain all option Greeks and their interpretation
• I understand vol surface and smile/skew
• I know order book mechanics, market impact
• I read Hull without difficulty

*Portfolio + competitions:*

• 3+ GitHub projects with READMEs
• Minimum 1 - serious AI/LLM project (differentiator in 2026)
• Top 10% in at least one trading competition

*AI fluency:*

• I understand architectures like AlphaGPT
• I built at least 1 multi-agent project
• I know where AI errors on financial data

*Interview:*

• Mental math: 50+/min on Zetamac
• LeetCode medium in 20-30 minutes
• DP: confident on medium
• 10+ mock interviews completed

Where to start tonight

*Free resources you can open today and work through without paying a dollar. The other books mentioned inline in the article are bought standardly through Amazon or found in open access via university repositories.*

*Open-access textbooks:*

• Blitzstein & Hwang - Introduction to Probability(https://stat110.hsites.harvard.edu/) - Harvard
• Hastie, Tibshirani, Friedman - Elements of Statistical Learning(https://hastie.su.domains/ElemStatLearn/) - Stanford
• Boyd & Vandenberghe - Convex Optimization(https://web.stanford.edu/~boyd/cvxbook/) - Stanford

*Courses:*

• Harvard Stat 110(https://projects.iq.harvard.edu/stat110) - probability, full YouTube playlist(https://www.youtube.com/playlist?list=PL2SOU6wwxB0uwwH80KTQ6ht66KWxbzTIo)
• MIT 18.06 (Strang)(https://ocw.mit.edu/courses/18-06-linear-algebra-spring-2010/) - linear algebra
• Karpathy - Neural Networks: Zero to Hero(https://karpathy.ai/zero-to-hero.html) - deep learning from scratch to GPT
• 3Blue1Brown - Essence of Linear Algebra(https://www.3blue1brown.com/topics/linear-algebra) - visual intuition

*Trainers:*

• Zetamac(https://arithmetic.zetamac.com/) - mental math (target 50+/min)
• Brainstellar(https://brainstellar.com/) - probability puzzles at interview difficulty
• LeetCode(https://leetcode.com/) - coding interviews
• Kaggle(https://www.kaggle.com/) - practice on real data

*Trading competitions (direct path to recruiters):*

• Jane Street Kaggle(https://www.kaggle.com/c/jane-street-real-time-market-data-forecasting) - $100k prize pool
• IMC Prosperity(https://prosperity.imc.com/) - annual trading game
• Optiver competitions(https://optiver.com/working-at-optiver/career-opportunities/)
• Citadel Datathon(https://www.citadel.com/careers/students/datathon/) - winners go into interview pipeline
• WorldQuant BRAIN(https://platform.worldquantbrain.com/) - cash for alpha signals

Final

The cycle goes like this. Citadel migrates to Google Cloud, other funds see the 96x backtest speedup and start migrating. Man Group launches AlphaGPT, competitors build their own multi-agent systems. Bridgewater AIA Labs trades client money via LLM workflow, setting precedent for the entire macro industry.

This is a profession that 10 years ago was accessible only to PhDs from MIT. 5 years ago it opened to strong masters. In 2026 it opens to STEM bachelors with the right preparation.

The bar is higher than ever: now you need both the classical quant stack and AI/LLM. The pay reflects this - $336k-$642k entry-level at Citadel comes from public H1B data, not marketing.

*AI opened the doors. The same AI makes the speedrun viable. An hour with Claude on Shreve gives more understanding than a week alone. Mock interviews are available any time. Backtest code review takes a minute. Reproducing an academic paper takes an evening instead of a week.*

The speedrun doesn't make the profession easy. Math still requires months of work. Mental math still needs to be drilled to 50+/min. Trading competitions still filter out 99% of participants. P&L is still the only judge. What AI does is remove the dead time between effort and feedback - the part of traditional learning where most people gave up.

For the first time in history, the path from zero to Jane Street is technically possible for someone without an MIT diploma. The condition: walk all 7 steps without skipping, and use AI as a real acceleration tool rather than a replacement for your own thinking.

Sources

*Industry transformation data (Part 2):*

• Man Group - What AI Can (and Can't Yet) Do for Alpha(https://www.man.com/insights/what-ai-can-do-for-alpha) - official technical post on AlphaGPT
• Man Group - A Trend Following Deep Dive: AI, Agents and Trend(https://www.man.com/insights/ai-agents-trend) - Alpha Assistant architecture details
• Bloomberg - Man Group Says Agentic AI Is Now Devising Quant Trading Signals(https://www.bloomberg.com/news/articles/2025-07-10/man-group-says-agentic-ai-is-now-devising-quant-trading-signals) (July 2025)
• Hedgeweek - Man Group deploys agentic AI for quant signal discovery(https://www.hedgeweek.com/man-group-deploys-agentic-ai-for-quant-signal-discovery/)
• AI Street - Inside Man Group's AlphaGPT(https://www.ai-street.co/p/inside-man-group-s-alphagpt) - interview with Senior PM Ziang Fang
• Bridgewater - research insights(https://www.bridgewater.com/research-and-insights) - AIA Labs publications

*Salary data:*

• H1B Salary Database(https://h1bdata.info/) - public hedge fund and prop shop compensation data
• Selby Jennings - Compensation Reports(https://www.selbyjennings.com/) - industry surveys

*Recruiting and interview data:*

• QuantNet - MFE Rankings & Forum(https://quantnet.com/) - program rankings and discussions
• Wall Street Oasis(https://www.wallstreetoasis.com/) - fund process discussions
• Jane Street - Tech talks (YouTube)(https://www.youtube.com/@janestreet) - official technical lectures