Foundation
Everyone starts hereMath and statistics
Probability, linear algebra, calculus, and statistics are the grammar of every model in finance and AI. Skip these and you are guessing.
- 3Blue1Brown Free YouTube · visual linear algebra and calculus, best in class
- StatQuest Free YouTube · statistics explained at a pace that actually sticks
- MIT OCW 18.06 Free · Gilbert Strang's linear algebra lectures, canonical
Python for data
Python is the working language of AI, quant, and data science. NumPy and pandas are load-bearing; learn them before libraries abstract them away.
- CS50P Free · Harvard's intro to Python, rigorous, self-paced
- freeCodeCamp Python (4h) Free YouTube · single-video crash course that actually covers the essentials
- Python for Data Analysis Free · Wes McKinney (pandas author), available free online
Track A: Quant / financial engineering
Stochastic everythingStochastic calculus
Brownian motion and Ito's lemma are the foundation of every derivative pricing model. Dense, but unavoidable if you are serious about this track.
- MIT Stochastic Processes (OCW) Free Lectures · rigorous graduate-level starting point
- Mathematical Finance (YouTube) Free · worked examples through SDE basics and Girsanov
Derivatives pricing
Black-Scholes is not the end; it is the beginning. Understanding why it works and where it breaks is what separates users from practitioners.
- Financial Engineering (Coursera) Audit Free · Columbia / Martin Haugh, covers BSM, numerics, calibration
- Black-Scholes derivation (YouTube) Free · clean 20-min walkthrough of the PDE from first principles
Fixed income and rates
Duration, convexity, and term-structure models underlie a large share of real risk capital. Bonds are less exciting and more consequential than options.
- MIT 15.401 Finance Theory (OCW) Free · covers bond math, duration, and yield curve basics
- Finance with Emily (YouTube) Free · accessible fixed income concepts with real examples
Market, credit, and model risk
VaR, expected shortfall, credit spreads, and model validation are the professional language of risk. Know the distinction between types before mixing them up in a pitch.
- BioniC Quant (YouTube) Free · risk metrics, VaR, and quantitative risk management topics
- GARP FRM Study Material Free · the FRM curriculum is public and covers all three risk types
Time series analysis
ARIMA, cointegration, and volatility models are the workhorses of financial modeling. Most "AI predictions" in finance are really time series econometrics with better PR.
- Ritvikmath Time Series (YouTube) Free · best free resource on ARIMA and the classics, bar none
- Forecasting: Principles and Practice Free · Hyndman and Athanasopoulos, free online, comprehensive
Track B: AI / machine learning
Models that learnML fundamentals
Gradient descent, overfitting, bias-variance tradeoff, cross-validation. The ideas that every other concept in this track depends on.
- Andrew Ng ML Specialization Audit Free · the canonical introduction, updated for modern frameworks
- StatQuest Free YouTube · covers decision trees, random forests, gradient boosting with unusual clarity
Deep learning
Neural networks, backpropagation, CNNs, and the architecture decisions that separate toy from production. Understanding this layer lets you reason about what LLMs actually do.
- fast.ai Practical Deep Learning Free · top-down, code-first, deliberately non-academic
- 3Blue1Brown: Neural Networks Free YouTube · visual intuition for what gradients and backprop are actually doing
NLP and LLMs
Transformers, attention, and how language models work from the inside. The difference between using an API and understanding the architecture is the difference between a user and a builder.
- Andrej Karpathy: Zero to Hero Free · builds a GPT from scratch in numpy, then PyTorch. Nothing better exists at this depth.
- 3Blue1Brown: Transformers Free YouTube · the attention mechanism explained visually
AI agents and tool use
Agents are models that act in the world: they call tools, maintain state, and loop until a task completes. This is where the capability ceiling sits right now.
- AI Agents in LangGraph (DeepLearning.AI) Free · short course, practical, covers the ReAct loop and tool calling patterns
- Anthropic Agent Docs Free · the reference implementation from the team that publishes the best thinking on agent design
Track C: Data science in finance
Applied, production-mindedSQL and data wrangling
Every finance data science job starts with SQL and messy datasets. Master joins, window functions, and pandas before touching a model.
- Mode SQL Tutorial Free · browser-based, goes from basics to window functions cleanly
- freeCodeCamp SQL (4h) Free YouTube · thorough single-video course; good for building muscle memory
ML applied to finance
Credit scoring, fraud detection, return prediction, and portfolio construction with ML. Finance requires calibrated uncertainty and explainability that generic ML courses skip.
- ML for Algorithmic Trading (GitHub) Free · Stefan Jansen's book repo with full notebooks; the most practical ML-in-finance resource available free
- ML for Finance (YouTube overview) Free · 90-min conceptual overview covering the domain-specific pitfalls
Forecasting in practice
Revenue forecasting, demand planning, and cash flow projection with modern tooling. The gap between a good model and a trusted model is calibration and honest uncertainty bounds.
- Nixtla StatsForecast (docs) Free · best open-source forecasting library right now; documentation reads like a textbook
- Forecasting: Principles and Practice Free · referenced again because it genuinely covers applied forecasting better than most paid courses
MLOps basics
Model deployment, monitoring, versioning, and retraining loops. In finance, a model that cannot be audited or redeployed quickly is a liability. This is the production gap most DS courses ignore.
- Made With ML Free · Goku Mohandas's MLOps curriculum, practical and well-maintained
- MLOps overview (YouTube) Free · 2-hour walkthrough covering CI/CD for ML, monitoring, and feature stores
Visualization and communication
A finance audience does not read model cards. Communicating uncertainty, trends, and model outputs clearly is a professional skill as real as the modeling itself.
- Storytelling with Data (chart guide) Free · Cole Knaflic's chart chooser; the reference standard for data communication
- Storytelling with Data (YouTube) Free · applied examples on making data legible to non-technical stakeholders
Where the tracks meet
AI in financeThis is where the quant foundation, the ML toolset, and the data science workflow converge. Sarthak's focus: end-to-end AI in finance, with human-in-the-loop validation and governance that can survive regulatory scrutiny.
Agentic AI in finance
Agents that read filings, run models, reconcile data, and escalate exceptions are real and running in production. The architecture decisions that make them trustworthy with money are not obvious.
- Building effective agents (Anthropic) Free · the clearest public writing on agent architecture trade-offs
- skgpt · a working example of an agent deployed on this site; source patterns visible in the public repo
Model risk for AI systems
SR 11-7 was written for statistical models. Applying its validation logic to neural networks and agents requires new thinking. This is the most underserved skill gap in financial AI right now.
- SR 11-7 Guidance (Federal Reserve) Free · read the primary source; the document that defines the vocabulary of model risk in US finance
- Model risk + AI (YouTube search) Free · no single canonical resource yet; the field is moving; search for recent practitioner talks
Human-in-the-loop design
Automation that cannot hand off cleanly to a human is a liability, not an asset. HITL is not a concession; it is an architecture choice that keeps the system auditable and recoverable.
- Google PAIR Guidebook Free · practitioner guide to human-centered AI design, grounded in real deployment experience
- PAI: Human Oversight in Finance Free · focused specifically on automated decision systems in financial services
AI governance and compliance
The regulatory environment for AI in financial services is moving fast. Model inventories, audit trails, explainability requirements, and the EU AI Act's high-risk classification for credit systems are shaping what you can build and how.
- NIST AI Risk Management Framework Free · the US government reference standard for AI governance; readable and practical
- EU AI Act (plain English) Free · high-risk system classification, transparency requirements, and what it means for finance AI
Resources listed are free or free-to-audit unless noted otherwise. None are sponsored. All views are my own and not those of my employer. To suggest a resource, reach out via email.