What you'll walk away with
Six outcomes this course is actually built around — not a syllabus dump.
Prompt & context design
Structure prompts and context windows that hold up outside a demo.
Production RAG
Chunking, embeddings, retrieval and evaluation — not just a vector DB tutorial.
Agents & tool use
Multi-step agents that call tools, retry, and fail predictably.
Fine-tuning basics
When to fine-tune vs. prompt, and how to run it on a budget.
Evaluation & guardrails
Build eval sets, catch regressions, and add safety checks before shipping.
Ship a capstone
Deploy a full LLM app with monitoring — your portfolio centerpiece.
Your learning path
Five stages, one flowing track — click a stage in the curriculum below and watch it light up here.
Curriculum
5 modules. Click any module to expand it — and to highlight its stage on the roadmap above.
- How transformers and tokenization actually work
- Reading model cards, context limits & pricing tradeoffs
- Setting up your dev environment and API access
- Lab: build your first API-driven chat script
- Structured prompting patterns and few-shot design
- Context window budgeting for long documents
- Output formatting, function calling & structured outputs
- Lab: build a document Q&A assistant
- Chunking strategies and embedding model selection
- Vector databases: indexing, filtering, hybrid search
- Evaluating retrieval quality before it hits production
- Lab: production RAG pipeline over real documents
- Multi-step planning and tool-calling loops
- Guardrails, retries and failure handling
- Multi-agent coordination patterns
- Lab: build an autonomous research agent
- When to fine-tune vs. prompt or retrieve
- Building an eval harness to catch regressions
- Deployment, monitoring & cost tracking
- Capstone: ship and demo your own LLM product
Weekly schedule
Live sessions run evenings IST. Filter by day, or just watch — today's slot highlights itself.
Your instructor
Riya Kapoor
Ex-Applied Scientist, building LLM products since GPT-3Riya has shipped RAG and agent systems used by teams at two Series-B startups, and has mentored 400+ engineers through applied AI cohorts.