Skip to content

Latest commit

 

History

History
241 lines (175 loc) · 11.6 KB

PROGRESS_TRACKER.md

File metadata and controls

241 lines (175 loc) · 11.6 KB

📊 My 90 Days of AI Engineering Tracker

Curated by: Ahmed Raza Khan • theahmedraza.com

Use this document to track your personal progress as you advance through the 90 days. Fork this repo, edit this file, commit your updates daily, and share your journey in public using #90DaysOfAIEngineering!


👤 Learner Profile


📈 Overall Progress Bar

Total Progress: [ 0 / 90 Days Completed - 0% ]

[░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░] 0%

🗓️ Phase 1: Foundations & Mechanics (Days 1–28)

Week 1 & 2: Python, Git and APIs (Days 1–14)

  • Goal: Command-line LLM summarizer with async HTTP and clean error handling.

  • Project Repo: Link to your Week 1-2 Project

  • Day 01: Python data structures (lists, dicts, sets, comprehensions).

  • Day 02: Python functions, type hinting, and OOP basics.

  • Day 03: Virtual environments (venv), pip, and package management.

  • Day 04: Git setup, staging, committing, and GitHub repository creation.

  • Day 05: Git branching, merging, and pull request workflows.

  • Day 06: Environment variables, .env, and secret hygiene.

  • Day 07: HTTP protocol basics, status codes, and JSON structure.

  • Day 08: Consuming REST APIs with httpx or requests.

  • Day 09: Async Python syntax: async, await, and asyncio.gather().

  • Day 10: Connecting to an LLM API (OpenAI, Anthropic, Gemini, or local Ollama).

  • Day 11: Building CLI arguments with argparse or click.

  • Day 12: Streaming LLM output in the terminal.

  • Day 13: Error handling: API rate limits, timeouts, and network drops.

  • Day 14: Project Ship Day: Finalize CLI summarizer, write README, and push to GitHub!


Week 3: How LLMs Work (Days 15–21)

  • Goal: Chat CLI with conversation memory, token counter, and live cost calculator.

  • Project Repo: Link to your Week 3 Project

  • Day 15: Neural network fundamentals and transformer architecture high-level intuition.

  • Day 16: Attention mechanisms: Self-attention and Multi-Head Attention.

  • Day 17: Tokenization: Byte-Pair Encoding (BPE) and inspecting tokens with tiktoken.

  • Day 18: Sampling hyperparameters: Temperature, Top-P, Top-K, and repetition penalties.

  • Day 19: Managing conversation history & sliding context window truncation.

  • Day 20: Calculating live token consumption and estimating dollar cost per prompt.

  • Day 21: Project Ship Day: Finalize Token & Cost Chat CLI and commit code!


Week 4: Prompting and Structured Outputs (Days 22–28)

  • Goal: Resilient data extractor converting unstructured text into validated Pydantic JSON.

  • Project Repo: Link to your Week 4 Project

  • Day 22: System prompts, role prompting, and delimiter techniques.

  • Day 23: Few-shot prompting and chain-of-thought (CoT) reasoning.

  • Day 24: JSON mode and native provider structured outputs (OpenAI / Anthropic tools).

  • Day 25: Defining schemas and field validations with Pydantic v2.

  • Day 26: Building an automated retry loop when validation fails.

  • Day 27: Testing against 20+ edge-case inputs (broken text, missing fields).

  • Day 28: Project Ship Day: Benchmark pass rate and ship your structured extractor!


🗓️ Phase 2: Full-Stack AI Services (Days 29–56)

Week 5: Backend with FastAPI (Days 29–35)

  • Goal: Production FastAPI service with SSE streaming chat, extractor endpoint, and CI.

  • Project Repo: Link to your Week 5 Project

  • Day 29: FastAPI setup, path operations, and dependency injection.

  • Day 30: Exposing the structured extractor as a POST endpoint.

  • Day 31: Implementing Server-Sent Events (SSE) for real-time token streaming.

  • Day 32: Adding API key authentication header and basic rate limiting.

  • Day 33: Writing unit and integration tests with pytest and httpx.AsyncClient.

  • Day 34: Setting up GitHub Actions CI workflow to run test suite on push.

  • Day 35: Project Ship Day: Verify single-command startup and green CI badge!


Week 6: Embeddings and Vector Search (Days 36–42)

  • Goal: Semantic search over local documents comparing vector search vs. keyword search.

  • Project Repo: Link to your Week 6 Project

  • Day 36: What are embeddings? Dense vector representations and semantic space.

  • Day 37: Similarity metrics: Cosine similarity vs. Dot product vs. L2 distance.

  • Day 38: Generating embeddings with local models (sentence-transformers) or APIs.

  • Day 39: Setting up a local vector store (ChromaDB or sqlite-vec).

  • Day 40: Indexing Markdown, text, and PDF documents with metadata.

  • Day 41: Benchmarking queries: Where semantic wins vs. where keyword search wins.

  • Day 42: Project Ship Day: Finalize semantic search CLI and document findings!


Weeks 7 & 8: Production RAG (Days 43–56)

  • Goal: End-to-end RAG system with citations, hybrid retrieval, and zero-hallucination guardrails.

  • Project Repo: Link to your RAG Project

  • Day 43: RAG architecture overview: Ingestion, Chunking, Retrieval, Generation.

  • Day 44: Chunking strategies: Fixed character vs. Recursive vs. Semantic chunking.

  • Day 45: Metadata tagging: Storing document source, section, and page numbers.

  • Day 46: BM25 keyword search implementation.

  • Day 47: Hybrid search: Combining dense vector similarity + sparse BM25 (Reciprocal Rank Fusion).

  • Day 48: Cross-encoder reranking (Cohere Rerank or FlashRank).

  • Day 49: Constructing grounded prompts with strict hallucination defense.

  • Day 50: Formatting structured answers with exact source citations ([Doc X, p. Y]).

  • Day 51: Handling out-of-domain questions ("I don't know" verification).

  • Day 52: Multi-query expansion and sub-question generation.

  • Day 53: Caching frequent queries and retrieval results.

  • Day 54: Evaluating chunking strategy performance side-by-side.

  • Day 55: Cleaning up API endpoints and adding CLI search interface.

  • Day 56: Project Ship Day: Ship production RAG repository with evaluation notes!


🗓️ Phase 3: Production, Evals & Agents (Days 57–84)

Week 9: LLM Evals & Reliability (Days 57–63)

  • Goal: Automated evaluation harness with golden dataset, LLM-as-a-judge, and CI gating.

  • Project Repo: Link to your Evals Project

  • Day 57: Why evals matter: Moving beyond "it looks good to me" vibe testing.

  • Day 58: Curating a 30+ question golden dataset with ground truth references.

  • Day 59: Defining evaluation metrics: Faithfulness, Context Precision, Answer Relevance.

  • Day 60: Implementing an LLM-as-a-judge evaluator prompt with rubric scoring.

  • Day 61: Identifying judge bias and reliability failure modes.

  • Day 62: Building an automated regression test script comparing prompt versions.

  • Day 63: Project Ship Day: Integrate eval runner into GitHub Actions CI!


Week 10: Tool Use and Agents (Days 64–70)

  • Goal: Autonomous agent with 3 tools, step limits, reasoning loop, and full audit logs.

  • Project Repo: Link to your Agent Project

  • Day 64: Function calling mechanics and tool JSON schemas.

  • Day 65: ReAct pattern: Reason → Act → Observe cycles.

  • Day 66: Building 3 practical tools: Web Search, Python Calculator, File Reader.

  • Day 67: Enforcing hard step limits and cycle detection to prevent infinite loops.

  • Day 68: Adding human approval prompts for risky tool executions.

  • Day 69: Structured logging of reasoning traces, tool inputs, and tool outputs.

  • Day 70: Project Ship Day: Demonstrate agent vs. plain prompt comparison!


Week 11: MCP and AI Coding Workflows (Days 71–77)

  • Goal: Custom Model Context Protocol (MCP) server wrapping an external API.

  • Project Repo: Link to your MCP Project

  • Day 71: Understanding Model Context Protocol (MCP) architecture & JSON-RPC.

  • Day 72: Inspecting existing MCP servers and tools in developer environments.

  • Day 73: Initializing an MCP server using the official Python SDK.

  • Day 74: Defining MCP tools with input schemas and typed arguments.

  • Day 75: Connecting your custom MCP server to Claude Desktop or an AI client.

  • Day 76: Writing unit tests for your MCP tools and error responses.

  • Day 77: Project Ship Day: Publish your MCP server with complete documentation!


Week 12: Deploy, Secure and Monitor (Days 78–84)

  • Goal: Dockerized deployment with rate limits, OWASP Top 10 defense, and tracing.

  • Project Repo: Link to your Deployment Project

  • Day 78: Writing an optimized multi-stage Dockerfile for FastAPI + AI dependencies.

  • Day 79: Docker Compose for local service orchestration (API + Vector DB).

  • Day 80: OWASP Top 10 for LLM: Testing prompt injection and jailbreaks.

  • Day 81: Setting up rate limiting, API key quotas, and cost budgets.

  • Day 82: Instrumentation: Request tracing, latency monitoring, and token logging.

  • Day 83: Deploying container to a cloud host (Render, Fly.io, or AWS Bedrock).

  • Day 84: Project Ship Day: Live public URL working end-to-end!


🗓️ Phase 4: Capstone & Career (Days 85–95)

Week 13: Capstone AI Product (Days 85–90)

  • Goal: Design, build, evaluate, and publicly launch an original AI product.

  • Capstone Repo: Link to your Capstone Project

  • Live Demo URL: Link to Live Demo

  • Day 85: Capstone product scoping: Defining the problem, user persona, and tech stack.

  • Day 86: Backend architecture: Ingestion, RAG/Agent pipeline, and API endpoints.

  • Day 87: Frontend UI: Clean interactive interface (Streamlit, Next.js, or HTML/JS).

  • Day 88: Running comprehensive evals and recording benchmark metrics.

  • Day 89: Writing the flagship README: Architecture diagram, demo GIF, and post-mortem.

  • Day 90: 🚀 LAUNCH DAY: Deploy live demo, publish on GitHub, and announce publicly!


Bonus: Portfolio, Resume & Applying (Days 91–95)

  • Day 91: Pin top 3 repositories on GitHub profile with clean cards and descriptions.
  • Day 92: Write a 1-page AI Engineer resume linking directly to live demos and evals.
  • Day 93: Publish a deep-dive technical article detailing lessons learned and tradeoffs.
  • Day 94: Submit an open-source pull request to an AI library you used.
  • Day 95: Apply to 10+ target AI Engineering roles and celebrate completing the 90 days! 🎉

📝 Daily Notes & Log Template

Copy and paste this section below for your daily journal:

Day XX (YYYY-MM-DD)

  • What I learned today:
  • What I built today:
  • Challenges faced & how I solved them:
  • Commit hash / PR link:
  • Tweet / LinkedIn post link: