Open-source quant Python stack

A short social thread this week collected a compact map of the open‑source Python tools quants are actually using, naming 17 libraries and pointing readers to ready‑to‑run notebooks for learning and backtesting. (x.com 1) (x.com 2) What the thread shows at a glance is not novelty but assembly: a handful of specialist projects that solve discrete problems in pricing, volatility modeling, portfolio construction, and execution, now available as plug‑and‑play Python packages. Goldman Sachs’ gs‑quant packages market data, instrument models and common analytics into a single client library so a researcher can call a standardized derivatives pricer instead of re‑implementing Black‑Scholes or a swap valuation every time. (github.com) Google’s tf‑quant‑finance supplies high‑performance implementations of option and rate models built on TensorFlow so model calibration can use GPUs and automatic differentiation instead of slow, hand‑tuned solvers. The project is archived, which means it’s stable code you can inspect and fork but no longer actively maintained by Google. (github.com) Other entries on the list are narrower but equally practical. pysabr implements the SABR family of models used to fit an “implied volatility smile” — the way option market prices imply different volatilities at different strikes — so a student can calibrate a swaption pricing model to real market quotes. (github.com) FinancePy collects pricing and risk routines for equities, rates, FX and credit, giving you a playground of instrument classes and numerical methods without starting from scratch. (github.com) Those libraries are useful on their own, but the social thread’s second nudge was toward pedagogy: Quantopian’s lecture notebooks, preserved and mirrored on GitHub, give a ready syllabus of quantitative techniques in executable form. The notebooks walk from basic statistics and pandas through regression, principal components, risk‑constrained portfolio optimization and practical backtests that produce the same plots and tables you’d expect in a research note. For a student building a portfolio project, that’s a set of lecture slides you can run, modify and re‑use as a reproducible experiment. (github.com) (gist.github.com) The practical upshot is concrete. If you want to demonstrate econometric chops, you can take a Quantopian notebook that teaches covariance estimation, plug in pandas and statsmodels, swap in gs‑quant or FinancePy valuation calls for instrument returns, and produce a backtest that shows how estimation noise affects portfolio turnover. If you want to probe derivatives work, you can use pysabr to fit a volatility surface and then feed that surface into a hedging simulation. The code paths are short: install a package, open a Jupyter notebook, run a calibration cell, and you have reproducible figures and tables to cite in a research portfolio. The thread also shows the division of labor in modern quant work: large firms contribute production‑grade toolkits, research groups contribute algorithmic primitives, and the community wraps them into teaching material. That mix matters for hiring — it rewards someone who can stitch models, data and evaluation together in a reproducible notebook, not just recite theory. If you want to start a project tomorrow, pick one problem (e.g., covariance estimation for heavy‑tailed returns), clone the relevant Quantopian lecture notebook as your scaffold (gist.github.com), and use a focused library — gs‑quant for instrument analytics or pysabr for volatility work — to produce a one‑page research note with code, calibration and a short backtest. (github.com 1) (github.com 2) (github.com 3)