The books that shape how I build, trade, and reason.
I keep a tight shelf — math, machine learning, quantitative finance, distributed systems, and a few books that bend the way I think. Every title here earned its place because it changed something I ship.
I pick one title per quarter to read twice — once for the argument, once for the craft. This is the current one.
Re-reading
480 pp.
AI / MLDutton · 2024
Why Machines Learn
The Elegant Math Behind Modern AI
by Anil Ananthaswamy
A patient walk through the mathematics — vectors, gradients, kernels, manifolds — that quietly powers everything from perceptrons to modern transformers, told through the lives of the people who discovered it.
Why it matters to me
It changed how I read papers. After this book, every transformer diagram stops being a black box and becomes geometry: high-dimensional spaces being rotated, projected, and untangled. That intuition leaks into every system I now design — from retrieval pipelines to model evaluation harnesses.
“Machines do not learn by magic. They learn by doing geometry in spaces we cannot picture, but whose rules we can write down.”
Linear AlgebraCalculusGeometryHistory of AIIntuition
Filter by domain. Each card carries the argument, my honest takeaway, and what it changed about the way I work.
Quant Finance2018
Advances in Financial Machine Learning
Marcos López de Prado
Re-reading·400p
The most rigorous treatment of applying ML to financial data: meta-labeling, fractional differentiation, purged cross-validation, and why most published quant strategies are statistical illusions.
Takeaway
If you want to ship a strategy that survives contact with live capital, this is the textbook. Triple-barrier labeling and purged k-fold alone reshaped how I evaluate every alpha I touch.
A compact, almost monograph-style follow-up. Hierarchical risk parity, denoising of covariance matrices, optimal portfolio construction without inverting noise.
Takeaway
Reads like a senior practitioner whispering across a trading desk. The HRP chapter alone replaced two years of my prior intuition about portfolio construction.
An uncompromising attack on associational factor research and a serious case for causal inference — DAGs, do-calculus, and falsifiable hypotheses — as the future of investing.
Takeaway
Reframed how I think about every signal I build: correlation is the question, not the answer. I now reach for causal graphs before I reach for a regression.
The definitive map of distributed systems for working engineers — replication, consensus, stream processing, and the trade-offs nobody warns you about until production breaks at 3am.
Takeaway
Every system diagram I draw now starts from this book's vocabulary. It is the closest thing our field has to a shared language.
A practitioner's playbook for mean reversion, momentum, and arbitrage strategies, with explicit code examples and brutally honest discussion of capacity and decay.
Takeaway
Chan does what most quant authors will not: he tells you when his strategies stopped working and why. That intellectual honesty is rarer than alpha.
Mean ReversionMomentumArbitrage+1
On the shelf
Mental Models2007
The Black Swan
Nassim Nicholas Taleb
Read·480p
Taleb's argument that history is shaped by rare, unpredictable, high-impact events — and that our models, narratives, and institutions systematically blind us to them.
Takeaway
Permanently changed my relationship with risk. I now design systems assuming the failure mode I did not anticipate is the one that will occur.
RiskEpistemologyHeavy Tails+1
On the shelf
Mental Models2012
Antifragile
Nassim Nicholas Taleb
Read·544p
The follow-up that names a property most engineers have felt but never had a word for: systems that get stronger under stress, not merely robust to it.
Takeaway
Reframed how I architect: the question is not 'will this survive load?' but 'does load make it better?' Chaos engineering, canary deploys, and incremental rollouts all click harder after this read.
Systems ThinkingOptionalityConvexity
On the shelf
Quant Finance1999
Active Portfolio Management
Grinold · Kahn
Read·596p
The classical foundation of quantitative active management — information ratios, the fundamental law, alpha forecasting, and risk modeling.
Takeaway
Old, dense, and irreplaceable. The fundamental law of active management is the single most useful equation I learned in finance.
Information RatioAlphaRisk Models
On the shelf
Systems2019
The Pragmatic Programmer
Hunt · Thomas
Read·352p
Twenty years of craft compressed into pragmatic principles: orthogonality, tracer bullets, broken windows, the boy scout rule.
Takeaway
I re-read a chapter at random whenever I feel my code getting precious. It keeps the ego sanded down and the hands moving.
CraftPrinciplesPragmatism
On the shelf
AI / ML2009
The Elements of Statistical Learning
Hastie · Tibshirani · Friedman
Up next·745p
The statistician's view of the same territory as Goodfellow — penalized regression, additive models, boosting, and the bias-variance trade-off in full mathematical color.
Takeaway
The book I open when a model behaves weirdly and I suspect the issue is statistical, not architectural. A different lens on the same problems, and richer for it.
If any of the ideas on this shelf — quant ML, causal inference, distributed systems, or just careful engineering — map to something you want built, I'd like to hear about it.
