Topfolio
www.topfolio.in

AI Engineer - Work Experience

Full 8-week curriculum

2 Months1-2 hrs/day (more during the agent + capstone weeks)Advanced → Job-Ready₹6,999 one-time
◆ WHAT'S INSIDE

Everything you'll learn, build, and use.

9
skills
7
tools
5
portfolio projects
WHAT YOU NEED TO START
  • Comfortable Python - the only prerequisite
  • 1-2 hrs/day for the full 8 weeks
Everything else is taught inside — Everything LLM - no prior AI/ML needed, resources land with each week's task · RAG, agents, evals, deployment · LangGraph, MCP, pgvector. Resources land with each week's task.
SKILLS YOU'LL ADD
LLM API engineering (structured output, cost, latency)Prompt & context engineering you can measureEmbeddings, chunking & retrieval qualityRAG with real evals (precision/recall + faithfulness)Tool-calling agents with retries & fallbacksLangGraph state machines & observabilityMulti-agent orchestration & MCPProduction deployment (auth, rate limits, monitoring)Benchmarking & failure post-mortems
Each skill comes with a curated video course when the task needs it — how it works
TOOLS YOU'LL USE⚙ Python⚙ Claude / GPT / Gemini APIs⚙ LangGraph⚙ MCP⚙ pgvector⚙ FastAPI⚙ Docker
◆ PORTFOLIO PROJECTS YOU'LL SHIP
WK 1
Structured-extraction microservice, deployed
FastAPI + Docker - your first shipped LLM system.
WK 2-4
RAG system + eval report
2 chunking strategies × 2 embedding models, scored on the numbers.
WK 5-6
Tool-calling LangGraph agent
Retries, fallbacks, full trace logging + a 20+ question eval suite.
WK 7
Multi-agent / MCP system
Documented architecture + persistent memory.
WK 8
Live capstone agentic app
Auth, rate limits, monitoring, its own golden benchmark + a failure post-mortem.
◆ 8-WEEK BREAKDOWN

What you'll ship, week by week.

WEEK
1
LLM API Mechanics & Structured Output
◆ SHIPS
Week 1
↳You'll arrive with: Just comfortable Python — every LLM concept starts from zero here, with curated resources landing alongside each week's task.
WHAT YOU'LL DO
  • ▸Calling Claude / GPT / Gemini like an engineer
  • ▸Structured / typed output you can trust
  • ▸Tokens, cost & latency basics
  • ▸Ship a small extraction microservice
SKILLS YOU'LL ADD
Call Claude/GPT/Gemini like an engineerGet structured, typed output you can trustReason about tokens, cost & latency
Curated learning resources land with this week's task.
TOOLS⚙ Python⚙ Claude / GPT / Gemini APIs
◆ YOU SHIP THIS WEEK
Project 1 ships: a deployed structured-extraction microservice
Reviewed in Discord · written feedback within 24h
WEEK
2
Prompt & Context Engineering
◆ SHIPS
Week 2
↳You'll arrive with: A shipped microservice and working API instincts.
WHAT YOU'LL DO
  • ▸Context windows & what to put in them
  • ▸Prompt patterns that actually generalize
  • ▸Few-shot vs instructions vs tools
  • ▸Measuring a prompt instead of vibes
SKILLS YOU'LL ADD
Engineer context windows deliberatelyBuild prompt patterns that generalizeMeasure prompts instead of trusting vibes
Curated learning resources land with this week's task.
TOOLS⚙ Claude / GPT / Gemini APIs
◆ YOU SHIP THIS WEEK
Advances Project 2: the measurement habits your RAG evals will run on
Reviewed in Discord · written feedback within 24h
WEEK
3
Embeddings & Vector Stores
◆ SHIPS
Week 3
↳You'll arrive with: You can measure LLM output quality — retrieval starts here.
WHAT YOU'LL DO
  • ▸How embeddings work and fail
  • ▸Chunking strategies that matter
  • ▸Vector stores (pgvector) in practice
  • ▸Retrieval quality, not just 'it returns something'
SKILLS YOU'LL ADD
Choose chunking strategies that matterRun a vector store in practiceEvaluate retrieval quality on numbers
Curated learning resources land with this week's task.
TOOLS⚙ pgvector⚙ Python
◆ YOU SHIP THIS WEEK
Advances Project 2: the retrieval layer of your RAG system
Reviewed in Discord · written feedback within 24h
WEEK
4
RAG + Evals (Project 1)
◆ SHIPS
Week 4
↳You'll arrive with: A working retrieval layer from Week 3.
WHAT YOU'LL DO
  • ▸A real RAG pipeline end-to-end
  • ▸Citations & grounding
  • ▸An eval set: retrieval precision/recall + faithfulness
  • ▸Compare chunking × embedding choices on the numbers
SKILLS YOU'LL ADD
Build a RAG pipeline end-to-endGround answers with citationsScore retrieval precision/recall + faithfulness
Curated learning resources land with this week's task.
TOOLS⚙ pgvector⚙ Eval tooling
◆ YOU SHIP THIS WEEK
Project 2 ships: RAG system + eval report (2 chunking × 2 embedding choices, scored)
Reviewed in Discord · written feedback within 24h
WEEK
5
Tool-Calling Agents
◆ SHIPS
Week 5
↳You'll arrive with: A shipped, evaluated RAG system.
WHAT YOU'LL DO
  • ▸Tools, function calling & the agent loop
  • ▸Retries, fallbacks & loop prevention
  • ▸When an agent is the wrong answer
  • ▸A 20+ question eval suite for your agent
SKILLS YOU'LL ADD
Build tool-calling agent loopsDesign retries, fallbacks & loop preventionJudge when an agent is the wrong answer
Curated learning resources land with this week's task.
TOOLS⚙ Claude / GPT / Gemini APIs⚙ Python
◆ YOU SHIP THIS WEEK
Advances Project 3: the agent core + a 20+ question eval suite
Reviewed in Discord · written feedback within 24h
WEEK
6
LangGraph + Observability (Project 2)
◆ SHIPS
Week 6
↳You'll arrive with: A working agent loop with its eval suite.
WHAT YOU'LL DO
  • ▸State machines for reliable agents
  • ▸Full trace logging of every step
  • ▸Debugging agents from traces
  • ▸Cost & latency you can actually see
SKILLS YOU'LL ADD
Model agents as state machinesTrace and debug every agent stepSee cost & latency per run
Curated learning resources land with this week's task.
TOOLS⚙ LangGraph⚙ Tracing tooling
◆ YOU SHIP THIS WEEK
Project 3 ships: a LangGraph agent with retries, full trace logging + evals
Reviewed in Discord · written feedback within 24h
WEEK
7
Multi-Agent & MCP (Project 3)
◆ SHIPS
Week 7
↳You'll arrive with: One reliable, observable agent in production shape.
WHAT YOU'LL DO
  • ▸Multi-agent orchestration patterns
  • ▸Connecting tools via MCP
  • ▸Persistent memory (Postgres/pgvector)
  • ▸Documenting your architecture decisions
SKILLS YOU'LL ADD
Orchestrate multi-agent patternsConnect tools via MCPDesign persistent memory
Curated learning resources land with this week's task.
TOOLS⚙ MCP⚙ Postgres/pgvector
◆ YOU SHIP THIS WEEK
Project 4 ships: a multi-agent / MCP-connected system with documented architecture
Reviewed in Discord · written feedback within 24h
WEEK
8
Deployed Capstone - Benchmark + Post-Mortem
◆ SHIPS
Week 8
↳You'll arrive with: Four shipped AI systems behind you.
WHAT YOU'LL DO
  • ▸Deploy the agentic app live (auth, rate limits, monitoring)
  • ▸Build its own golden benchmark
  • ▸Write a failure post-mortem on a real bug you fixed
  • ▸Mock AI-engineer interview on your own system
SKILLS YOU'LL ADD
Deploy with auth, rate limits & monitoringBuild a golden benchmark for your own appWrite an honest failure post-mortem
Curated learning resources land with this week's task.
TOOLS⚙ FastAPI⚙ Docker
◆ YOU SHIP THIS WEEK
Project 5 ships: the live capstone agentic app + benchmark + post-mortem
Reviewed in Discord · written feedback within 24h
◆ HOW IT RUNS

Yes, you get videos. Yes, you get live time — but neither is the program; the work is. Real tasks with deadlines · a curated video course per skill, watched when the task needs it · AI as your pair-worker, reviewed so you understand every line · written review of every submission within 24 hours · a personal weekly 30-min 1:1. Full details →

₹6,999 one-time · pause anytime · lifetime access
See the program & enrollAsk Anuj on WhatsApp
AI Engineer - Work Experience - Full 8-Week Curriculum & Brochure