Programme Content
BUILD
AI Foundations for FDEs
GPT, Claude, Gemini, open-weights, transformer architecture at FDE level, the FDE mental model: AI in client environment vs demo.
BUILD
Python & Production Engineering
Production Python (async, type-safety), FastAPI, Pydantic, Git workflows, structured logging, Docker. The things that make code shippable.
DELIVER
Discovery & Problem Framing
Run discovery sessions, frame AI use cases, map stakeholders (CTO, CISO, business owner). Turn a problem into a buildable scope.
BUILD
No-Code AI Prototyping
Visual RAG, agents and automations in IIT Delhi's QInsights lab. Sandbox access for every learner.
BUILD
Prompting & LLM Control
Few-shot, chain-of-thought, ReAct, structured outputs, prompt-as-code, and the quirks of each model.
◆ MINI-CAPSTONE 1
Ship your first working AI tool. Built end-to-end, documented like an FDE — a clean one-pager.
2. Move from tools to systemsBUILD
RAG Architecture IVector databases (Pinecone, pgvector, Qdrant, Chroma), embeddings, chunking, hybrid search.
BUILD
RAG Architecture IIRe-ranking, query rewriting, citation handling, multi-tenant retrieval.
BUILD
Agentic Systems I & IIAgent patterns, tool / function calling, ReAct, planner-executor. MCP, LangGraph, CrewAI, agent debugging.
BUILD
LLM Evaluation & ObservabilityRAGAS, DeepEval, golden datasets, LLM-as-judge, regression testing, LangSmith / Langfuse, red-teaming.
DELIVER
PRDs, Solution Design & Trade-off CommsPRDs for AI products, solution-design docs, communicating trade-offs to non-technical stakeholders, scoping pilots vs production.
◆ MINI-CAPSTONE 2
Build production RAG on messy enterprise docs. Full eval pipeline + architecture write-up.
3. Production. Security. ScaleBUILD
Fine-Tuning & CustomisationWhen to fine-tune vs prompt vs RAG. SFT, LoRA / QLoRA, distillation, dataset prep, eval gates, cost-benefit math.
BUILD
Deploying AI in ProductionAWS Bedrock, Azure OpenAI, GCP Vertex. Serving frameworks, latency optimisation, caching, autoscaling.
BUILD
Cost, Latency & ScaleToken economics, caching layers, model routing, batching. What turns an unprofitable demo into a profitable product.
BUILD
Enterprise IntegrationCustomer data schemas, lineage, secrets, audit trails. SOC2, HIPAA, DPDP. On-prem and VPC deployments.
BUILD
Security & Responsible AIOWASP for LLMs, prompt-injection defence, hallucination handling, PII, bias audits, DPDP & EU AI Act.
DELIVER
Stakeholder AlignmentDelivery planning, iteration cycles, managing scope changes, running war rooms during incidents.
◆ MINI-CAPSTONE 3
Rebuild a generic demo for a bank or hospital. Architecture + prototype + risk register.
◆ MINI-CAPSTONE 4
Audit a live LLM app end-to-end. Audit report + remediation playbook + governance plan.
4. Real FDE work, end-to-endBUILD
Vertical PlaybooksDeep dives into named domains. Fraud detection, ticket routing, multilingual workflows, clinical-grade healthcare RAG.
BUILD
AI Product SenseSpotting patterns across deployments. When to build vs configure vs raise to product. The "eyes and ears in the field".
DELIVER
Outcome Communication & Demo CraftDemoing to a Fortune-500 CTO. Deployment docs customers actually use. Presenting outcomes, not features. Portfolio storytelling.
★ FINAL CAPSTONE
Choose your track. Build the prototype. Harden it. Demo it — defended live to an SME panel.
5. QINSIGHTS NO-CODE LAB ACCESSSee the principles. Then build them.
Every learner gets access to IIT Delhi's QInsights No-Code Lab, one of the world's very few visualisation labs. Principles of modern AI (RAG, agents, retrieval, evaluation) are taught visually first, with minimal setup effort, before you write production code.
QInsights
VISUAL IIT DELHI LAB
Visualise RAG retrieval flows
Trace agent decision paths
Prototype with minimal code
If you are getting ready to work inside the QInsights environment, tell me:
If you have a specific use case to prototype (e.g., customer support agent, document search)
The AI components you want to visualize first
If you want to discuss how to transition visual prototypes into production Python code
Python
OpenAI & Claude APIs
LangGraph
LangSmith
Pinecone
Weaviate
Docker
Kubernetes
AWS
Azure
GCP
CI / CD pipelines
Eval frameworks
SQL & vector DBs
