Why I Built a Day-by-Day Plan — Not Just Another Roadmap

Most AI learning roadmaps tell you what to cover. They give you a list of topics, maybe some course links, and leave you to figure out the rest. After 13 years in software and XR development — and after going through the painful process of adding AI to my own projects starting in 2020 — I knew that was not enough.

The problem is not knowing what to learn. The problem is knowing exactly what to do on Day 1, Day 15, Day 47. What to build, what dataset to use, how long to spend on it. This plan solves that. Every single one of the 80 days has four things: what to learn, what to build, a free dataset or resource, and a writing prompt to force you to think about what you did.

Why BFSI — Banks, Asset Managers, Fintech

I deliberately tuned this plan for BFSI — Banking, Financial Services, and Insurance. Here is why. BFSI companies are among the largest employers of AI Architects in India and globally. The domain is complex enough to require real expertise. And the datasets — fraud detection, credit risk, SEC filings, RBI circulars, regulatory compliance documents — are all publicly available for free.

More importantly, BFSI projects look serious on a portfolio. A fraud detection model with SHAP explainability, a RAG system built on regulatory filings, an agentic AI that writes investment memos — these tell a hiring manager something very different from a generic house-price predictor or a Wikipedia chatbot. Domain specificity is a differentiator.

How the 80 Days Are Structured

Phase 1 (Days 1–10) starts with Python and math — but immediately working on financial data. You pull stock prices on Day 1. Phase 2 (Days 11–22) builds a fraud classifier and credit risk predictor — real BFSI use cases, not toy problems. Phase 3 (Days 23–38) goes deep into transformers and fine-tuning, building a financial news sentiment model. Phase 4 (Days 39–46) covers embeddings and vector search using RBI and SEC documents as the corpus. Phase 5 (Days 47–54) builds a production RAG system for financial document Q&A. Phase 6 (Days 55–62) moves into agentic AI — multi-agent systems for financial research. Phase 7 (Days 63–70) covers MLOps, deployment, monitoring, and explainability. Phase 8 (Days 71–80) is pure architecture — system design, ADRs, case studies, and interview prep.

How to Use This Page

Click any day card to see the full detail — what to learn, what to build, which free dataset to use, and the writing prompt for that day. Filter by phase using the phase buttons, or filter by library to see which day each tool is introduced. I am also publishing a dedicated blog post for each day as I go through the plan myself — those posts link back here and go much deeper on each topic.

The full curriculum for all 80 days is already live in the interactive plan below. Use the phase buttons to jump to any phase, or use the library filter to see exactly which day each tool is introduced. The detailed daily write-ups are published on the blog as I work through each day — visit the blog to follow along.