量化交易员从0到年薪65万美元的进阶路线图

How to become a quant from 0 to $650k/year (PLAYBOOK):

by the end, you'll know how to:

• build the exact math foundation top firms actually test - ship real projects that separate you from every other applicant - land your first credential and use it to climb the staircase - crack interviews at Citadel, Jane Street and prop shops

btw i was quantitative backend developer at Liquid Hedge Fund based in Dubai for almost a year, building systematic trading strategies across perp dexes and prediction markets (right now on a break, working on my own projects)

everything below is what actually got me here, in the order it actually worked

here's the full 5-stage roadmap:

Stage 1 (year 0): build the math foundation in the correct order

the biggest mistake is starting with finance textbooks or trading blogs

you don't need to know markets. you need to know math that markets run on

how to build it:

> probability first - Blitzstein & Hwang, free PDF from Harvard. chapters 1 to 6. conditional thinking is the entire skill > statistics second - Wasserman, All of Statistics. how to tell signal from noise > linear algebra third - Strang MIT 18.06, free at > OpenCourseWare. eigenvalues reveal what actually matters > convex optimization fourth - Boyd & Vandenberghe, free from Stanford. every portfolio problem reduces to this > stochastic calculus last - Shreve Vol 1 and 2. this is where Black-Scholes comes from

your first year should feel painful, slow, non-linear

most people quit here. that's why the field pays what it pays

metric to watch: can you derive Ito's Lemma from scratch on a whiteboard?

mistake to avoid: skipping layers because they feel abstract

Stage 2 (year 1): read the papers and implement each one from scratch

this is where I separated from every other backend developer trying to get in

the 5 papers that shaped every quant desk on Wall Street:

> Avellaneda & Stoikov 2008 (market making) > Engle & Granger 1987 (cointegration) > Heston 1993 (stochastic volatility) > Fama & French 2015 (five-factor model) > Cohen, Malloy, Pomorski 2012 (insider clusters)

read all five. twice. then implement each from scratch in Python. no frameworks

language stack that actually ships:

> Python for research, backtesting and 90% of quant work. pandas, polars, pytorch, cvxpy, statsmodels > Rust for new systematic infrastructure. this is what Hyperliquid, prediction market engines, and next-gen HFT stacks are built in > Go for execution infrastructure. broker connectivity, microservices, low-latency APIs > C++ only if you're targeting elite HFT engineering. Jane Street uses OCaml specifically for their trading stack

the mistake most self-taught quants make is picking one language and defending it, real desks run POLYGOT.

Python for research, Rust or Go for execution

what this gave me:

1: real understanding of how each model breaks 2: code on GitHub that recruiters actually respected 3: the ability to explain any of these in an interview without notes

Heston was the one that changed my thinking. once I understood why volatility itself follows a stochastic process, options started making sense

metric to watch: can you code Heston from a blank file in under 4 hours?

mistake to avoid: reading 20 papers superficially instead of 5 deeply (VERY IMP)

Stage 3 (year 2): ship 6 real projects with working backtests

this is where most aspiring quants die silently

the 6 projects that will land your first interview:

> pairs trading bot on 5 cointegrated equity pairs > options pricing engine with Heston implementation > Fama-French factor decomposition scanner > insider cluster detector using SEC EDGAR > mean reversion scanner for S&P 500 > cross-exchange arbitrage scanner for perp DEXs

every project needed: - clean GitHub repo with a proper README - working backtest with real historical data - honest documentation of what failed and why - one paragraph explaining the math in plain English

the honest documentation is what actually got me noticed. I wrote a section called "what broke" in every README. two recruiters told me later that was the reason they DM'd me

metric to watch: how many recruiters DM you after posting these

mistake to avoid: hiding the projects where the math didn't work

Stage 4 (year 3): first credential and the interview grind

credential path (what worked for me):

> cold emailed 40 PhD students. 3 responded. 1 took me on > TA'd a quantitative finance course at a local university > contributed to open-source quant libraries on GitHub > the title didn't matter. having real technical work to talk about mattered

interview prep: - the Green Book (Xinfeng Zhou). 200 real interview problems. spend 15 minutes attempting each before looking at hints - QuantGuide(dot)io for probability puzzles at interview difficulty - Brainstellar for brain teasers - Zetamac for mental math. target 50+ correct per minute before applying - LeetCode Blind 75 with focus on dynamic programming

the interview is a filter. it's not testing your knowledge, it's testing how you think under pressure with incomplete information

metric to watch: how quickly you can solve fresh problems out loud

mistake to avoid: memorizing solutions instead of understanding patterns

Stage 5 (year 3-4): compete publicly and climb the staircase

organic applications alone hit a ceiling. now you layer public competitions:

> Jane Street Kaggle ($100K prize + direct recruiter visibility) > WorldQuant BRAIN (paid alpha signals + institutional pipeline) > Citadel Datathon (winners fast-tracked to interview rounds)

the staircase (this is the part nobody tells you):

1: research lab or startup first 2: mid-tier prop shop second 3: elite fund like Citadel, Jane Street, or Two Sigma third

nobody jumps directly from zero to Jane Street. everyone climbs

attempting the vertical jump and failing doesn't mean the field is closed. it means you tried to skip stairs

metric to watch: interview conversion rate at each tier

mistake to avoid: applying to elite firms first, getting rejected and giving up

CONCLUSION

most aspiring quants die between the math foundation and the first credential because they think they need one more course, one more book, one more year

the truth is: the first 3 stages are slow, painful and non-linear

but they build the foundation that makes stages 4 and 5 feel like gravity

start with probability. read the 5 papers. ship 6 projects. then apply

no MIT. no Stanford. no finance degree required

Citadel entry level pays $336K to $642K. Jane Street average $1.4M. 5-year top prop $800K to $1.2M

the information asymmetry keeping most people out isn't intelligence. it's not knowing what the path looks like.

COMPLETE SUMMARY BELOW: