Quant Career Roadmap

Quant Career Roadmap Workflow

The full path from zero to a (often remote) quant job — no PhD required. Consult the quant-career-docs skill-graph for every formula, tool, and resource, and route hands-on work to the role specialists (quant_developer, quant_researcher, quant_trader, risk_analyst). The non-negotiable standard throughout is honest analysis: always report Sharpe, max drawdown, and failure regimes.

Steps

Step 1: math-foundations

Agent: quant_career_mentor

Months 1–4. Build comfort with linear algebra, calculus, probability & statistics (the core engine), and optimization via MIT OCW / Coursera "Mathematics for ML" / 3Blue1Brown. No university course required. Expected: math-readiness

Step 2: python-stack

Agent: quant_developer

Months 1–4. Learn the stack in order: NumPy, Pandas, Matplotlib/Plotly, SciPy, scikit-learn, yfinance. First-day exercise: pull 5y of AAPL and plot a 30-day rolling mean in a Jupyter notebook. Expected: python-fluency

Step 3: finance-fundamentals [depends_on: Step 1]

Agent: quant_career_mentor

Months 2–5. Markets 101 (price discovery, spread, order types), options & the Greeks + Black-Scholes, bonds & duration, hedge-fund types, and the three canonical edges (mean reversion, momentum, factor models). Expected: finance-vocabulary

Step 4: first-strategy-backtest [depends_on: Step 2, Step 3]

Agent: quant_developer

Months 3–6. Build a leak-free backtest in Backtrader/Zipline: SMA crossover first, then pairs trading with cointegration. Compute Sharpe (>1.0 decent, <0.5 redo). Audit for lookahead bias. Ship the first GitHub project. Expected: first-backtested-strategy

Step 5: ml-for-finance [depends_on: Step 4]

Agent: quant_researcher

Months 5–8. Regime detection with HMM, XGBoost/LightGBM on tabular features, FinBERT sentiment. Predict direction not price; treat ML as feature discovery and regime identification, not magic forecasting. Expected: ml-finance-capability

Step 6: portfolio-projects [depends_on: Step 4, Step 5]

Agent: quant_career_mentor

Months 6–12. Build at least 3 (ideally 5) of: pairs trading, Fama-French factor model, GARCH volatility forecasting, FinBERT sentiment alpha, S&P-direction ML classification. Enforce the GitHub rule: README (what + why), metrics (Sharpe/max-drawdown/CAGR vs benchmark), clean code, honest "what didn't work". Expected: github-portfolio

Step 7: certifications [depends_on: Step 3]

Agent: quant_career_mentor

Optional, runs alongside the portfolio. CQF, CFA Level 1, QuantInsti EPAT, Andrew Ng ML Specialization (do before financial ML), MIT OCW 18.S096. Pursue if budget and goals warrant. Expected: credentials

Step 8: job-search [depends_on: Step 6]

Agent: quant_career_mentor

Months 7–12. Channels: LinkedIn, eFinancialCareers, QuantConnect, Numerai. Target remote niches (crypto quant, fintech risk, AI+finance, consulting). Role-specific interview prep: LeetCode/system-design (QD), brainteasers + Zetamac (QR/QT). Expected: offers-and-interview-readiness

Step 9: kg-persist [depends_on: Step 8]

Agent: quant_career_mentor Tools: graph_write

Persist the learner's progress, completed projects, and metrics as typed nodes in the Knowledge Graph, linked to the quant role and project entities.

Output

• A phased, dependency-aware learning plan tracked in the KG
• A GitHub portfolio of 3–5 honestly-analyzed quant projects
• Interview-ready preparation and a realistic remote job-search plan

Execution

Run this workflow as a dependency-ordered DAG. Steps with no unmet depends_on run in parallel; dependents run after their prerequisites complete.

• Run first (in parallel): Step 1 — math-foundations; Step 2 — python-stack
• After level 0: Step 3 — finance-fundamentals
• After level 1: Step 4 — first-strategy-backtest; Step 7 — certifications
• After level 2: Step 5 — ml-for-finance
• After level 3: Step 6 — portfolio-projects
• After level 4: Step 8 — job-search
• After level 5: Step 9 — kg-persist

Execution: If graph-os is reachable, offload the whole DAG via graph_orchestrate action=execute_workflow (or the kg-delegate skill) for true parallel/swarm execution. Otherwise execute the steps natively in dependency order: run steps with no unmet depends_on in parallel, then their dependents.