Live Cohort · 8 Weekends · Limited Seats

Master AI Engineering

You can already ship software. But "add AI to it" now means RAG pipelines, agents, tool calling, evals, and guardrails, and most tutorials stop at a toy demo that falls apart in production. This 8-week live cohort is the hands-on path: you learn each building block, then build it the same session, week after week, until you ship a real capstone you can put in front of an interviewer.

📅
Duration
8 Weeks · Weekends
⏰
Timing
Sat & Sun · 9:00–11:30 AM IST
🚀
Next Batch
Oct 18
🎓
Outcome
Ship a capstone
Enroll: Chat on WhatsApp See the full syllabus ↓

Built for engineers moving into AI Engineer and Forward-Deployed Engineer roles · 8 hands-on weeks · 3 capstone tracks

Sound familiar?

Most engineers don't struggle with AI because they lack intelligence. They struggle because every tutorial stops at "hello world" and never shows the production reality.

"My RAG demo works on 5 docs, then dies in prod."It looked great in the notebook, but real corpora, bad chunks, and irrelevant retrievals crater the accuracy the moment users show up.

"I can call an LLM, but I can't build an agent."You know the API. You still can't wire up tool calling, a reasoning loop, and structured outputs into something that reliably gets work done.

"I glued a LangChain tutorial together and it breaks."When it fails you have no idea why, because you never learned evals, tracing, or how to reason about the context window.

"Everyone name-drops MCP, evals, guardrails."You keep hearing the terms in job descriptions but have never actually built one, so interviews expose the gap fast.

The fix isn't more videos. It's structured practice: learn one block, build it the same session on a real use case, then compose the blocks into agents and pipelines you can defend in an interview.

The learn → build method

Every concept is followed by hands-on code the same session. That's how it becomes a skill, not a bookmark.

1

Learn the building block

One core idea at a time, in a deliberate sequence: LLM basics, retrieval, agents, evals. Each week builds on the last.

2

Build it immediately

The same session, you write the code: a RAG pipeline, a tool-calling agent, an MCP connection, an eval suite. Theory turns into a working artifact while it's fresh.

3

Compose the blocks

As your toolkit grows, you combine them: advanced retrieval, multi-agent orchestration, guardrails and observability wrapped around real workloads.

4

Ship a capstone

The final weeks you scope, build, and present a production-shaped project end-to-end, with eval results and a demo you can show off.

The 8-week syllabus

Live sessions, Sat & Sun, 9:00–11:30 AM IST. Prerequisites: prior coding experience and a basic understanding of software systems. Tap any week to expand. Every week = learn the theory, then build it in a hands-on lab.

Week 1

Terminology, Prerequisites & the Model Landscape

The vocabulary and the provider tradeoffs before you build

Theory

  • Core LLM terminology: tokens, context window, parameters, attention
  • How LLMs are trained: pre-training, post-training, and inference at runtime
  • Embeddings & semantic similarity: turning text into vectors
  • Prompting basics: system, user, few-shot
  • Structured outputs and JSON formatting
  • Choosing a provider: frontier APIs (Anthropic, OpenAI, Google) vs open-weight models (Llama, Mistral, Qwen), cost, latency, data-residency tradeoffs New

Hands-On Coding

  • Make your first LLM API calls
  • Compute cosine similarity from scratch
LLMsEmbeddingsTokensPromptsAPI BasicsModel Selection
Week 2

RAG: Components & Architecture

The 5-stage pipeline that grounds an LLM in your data

Theory

  • Why RAG exists: knowledge cutoff and the hallucination problem
  • The 5-stage RAG pipeline: ingest, chunk, embed, index, retrieve
  • Chunking strategies and when to use each
  • Embedding models: choosing the right one for your use case
  • Vector databases & indexing algorithms (HNSW, IVF, PQ)

Hands-On Coding

  • Build an end-to-end RAG pipeline with LangChain + ChromaDB
ChunkingEmbeddingsVector DBHNSWRetrievalAugmentation
Week 3

Advanced RAG

Breaking through the accuracy ceiling of naive RAG

Theory

  • Why naive RAG hits accuracy ceilings
  • Query rewriting techniques: expansion, HyDE, multi-query
  • Cross-encoder reranking: LLM-quality scoring after initial retrieval
  • Metadata filtering and hybrid search (dense + sparse)
  • Evals specific to RAG: faithfulness, answer relevance

Hands-On Coding

  • Apply advanced retrieval techniques to your Week 2 pipeline
  • Benchmark the accuracy improvements with an eval suite
HyDERerankerQuery RewritingMulti-VectorSelf-RAG
Week 4

RAG Architectures & the RAG-vs-Long-Context Decision

Advanced patterns, and when not to build RAG at all

Theory

  • Common pitfalls in production RAG systems
  • GraphRAG: knowledge graphs as retrieval backends
  • KAG (Knowledge-Augmented Generation): structured KB + LLM
  • Agentic RAG: the LLM decides what to retrieve, when, and how
  • Multimodal RAG: retrieving across text, images, and tables
  • Decision framework: when RAG is right, when long-context stuffing beats it, when to skip retrieval entirely New

Hands-On Coding

  • Build a GraphRAG system on a real use case using Neo4j
  • Compare it against vanilla RAG
GraphRAGAgentic RAGKAGMultimodal RAGLightRAGLong Context
Week 5

Tool Calling, MCP & Single-Agent Systems

Turning an LLM into an agent that gets work done

Theory

  • What is an agent? LLM + tools + loop
  • LLM vs agent: how to tell a pipeline from a single API call
  • Tool / function calling: how the LLM triggers Python functions
  • Model Context Protocol (MCP): standardized tool/data connectivity, architecture, servers, clients, and why it's replacing bespoke integrations New
  • The ReAct pattern: Reasoning, Acting, Observing
  • Pydantic AI: type-safe agents with validated structured outputs

Hands-On Coding

  • Build a single agent with tools, structured outputs, and basic guardrails
  • Connect your agent to an MCP server
ReActTool CallingMCPPydanticAIPrompt EngineeringStructured Outputs
Week 6

Multi-Agent Systems

Orchestration, routing, and memory across agents

Theory

  • Why multi-agent? Parallelism, specialisation, separation of concerns
  • Agentic design patterns: Orchestrator-Worker, Routing
  • Problems unique to multi-agent: orchestration, information isolation, planning
  • Memory systems for agents: short-term and long-term
  • Cost and latency considerations in multi-agent systems

Hands-On Coding

  • Build a research + writer + critic multi-agent pipeline in LangGraph with routing
LangGraphMulti-AgentOrchestrationDesign PatternsRouting
Week 7

Context Engineering, Evaluation, Guardrails & Production

The densest week: making agents accurate, safe, and observable

Theory

  • Context engineering vs prompt engineering
  • What goes in the context: instructions, tools, history, retrieved docs
  • Quantitative evals (accuracy, F1, exact match, tool-call accuracy)
  • Qualitative evals with LLM-as-a-Judge: rubric design, single-answer grading
  • Prompt injection: attack patterns and defenses for agents with tool access New
  • Output validation and jailbreak resistance for user-facing systems New
  • Observability and tracing in production (Langfuse, LangSmith or equivalent) New
  • Serving patterns: streaming responses, async execution, latency budgets New

Hands-On Coding

  • Run a full eval suite on your Week 6 multi-agent pipeline (quantitative + LLM-as-a-Judge)
  • Red-team your pipeline and add input/output guardrails and tracing
Context EngineeringMemoryLLM-as-a-JudgeEvalsPrompt InjectionGuardrailsObservabilityServing
Week 8

Capstone Project

Scope it, build it, present it

Theory

  • AI engineering best practices: from prototype to production
  • Engineering decision framework: choosing RAG type, agent pattern, and stack
  • How to scope a capstone: MVP definition, what to cut, what to keep
  • Architecture reviews and 1:1 feedback on each project plan
  • Presenting AI systems in interviews: structure, demo flow, eval results

Hands-On Coding

  • Live capstone demos with peer Q&A
  • Eval walkthroughs and structured feedback
CapstoneArchitecture DesignPresentationPortfolio

Capstone project tracks

Pick one and ship it end-to-end. Each track is a production-shaped system, not a toy demo.

Track 1

FinQuery

Deterministic personal-finance AI agents with tool-calling ReAct, typed SQL execution, a hybrid-search knowledge base, and a reversible pseudonymization gate for 0% raw PII leakage to the LLM.

Track 2

Product Discovery Copilot

Root-cause attribution for product funnels using a cyclic, stateful multi-agent system with a dedicated "Falsifier" agent, Pydantic AI tools, typed SQL via DuckDB, and hybrid BM25 + cross-encoder reranking.

Track 3

GrantMatch

Transparent scholarship-eligibility agents combining retrieval over compliance PDFs, active web search, and structured JSON rulesets to generate cited eligibility reasoning.

What you'll walk away with

By week 8, "add AI to it" stops being vague and becomes a set of engineering decisions you can make and defend.

+Build a production RAG pipeline end-to-end: ingest, chunk, embed, index, retrieve, rerank, and eval.
+Know when not to use RAG, and when long-context or no retrieval is the better call.
+Build single and multi-agent systems with tool calling, the ReAct loop, and MCP.
+Add evals, guardrails, tracing, and serving so your agents are trustworthy in production.
+Ship a capstone project you can demo, with real eval results, for your portfolio.
+Interview confidently for AI Engineer and FDE roles, speaking the production vocabulary fluently.

Is this for you?

Perfect if you're…

  • A software engineer upskilling into AI Engineer or FDE roles
  • A backend / full-stack dev who wants real production AI skills
  • Comfortable coding, but new to RAG, agents, and evals
  • Tired of tutorials that stop at a toy "hello world RAG"
  • Wanting a portfolio-ready capstone, not just certificates

Not for you if…

  • You've never written code and want an intro-to-programming class
  • You want to train or fine-tune foundation models from scratch (this is applied AI engineering)
  • You're looking for a no-code tool tour
  • You can't commit ~3 hours per weekend for 8 weeks

Everything included

One price, full access, no upsells.

🎥

16 Live Sessions

Every Sat & Sun for 8 weeks, recorded if you miss one.

🧪

Hands-On Lab Every Week

You write real code each session, from first API call to full pipelines.

🏗️

3 Capstone Tracks

Choose FinQuery, Product Discovery Copilot, or GrantMatch and ship it.

🛡️

Production Playbooks

Evals, guardrails, prompt-injection defense, tracing, and serving patterns.

💬

Mentor Access

Direct doubt-clearing over the cohort group, never stay stuck.

🏆

Capstone Review

Architecture review and 1:1 feedback on your project plan and demo.

What learners say

"I'd shipped backend for years but froze on anything AI. Building a RAG pipeline and an agent the same session I learned them is what finally made it click."

RK
Rahul K.Backend SWE → AI Engineer

"The RAG-vs-long-context week alone saved me from over-engineering a pipeline my project didn't even need. Very production-minded."

TS
Tanvi S.Full-stack, 5 yrs

"MCP, evals, guardrails were just buzzwords to me before. Now I've actually built them and can talk through the tradeoffs in interviews."

AM
Ankit M.SDE-2, moving into FDE

"The capstone is the part that mattered. I walked out with a real project and eval numbers to show, not another certificate."

DN
Divya N.Platform engineer

Reserve your seat

Small cohort · personal attention · limited seats.

🔥 Next batch starts Oct 18 · seats fill fast, reach out early to lock your spot
Next Batch · Oct 18
₹39,999
One-time · all 8 weekends + recordings + labs
  • 16 live sessions (Sat & Sun, 9:00–11:30 AM IST)
  • Learn-then-build format with hands-on labs every week
  • 3 capstone tracks + 1:1 architecture review
  • Production playbooks: evals, guardrails, tracing, serving
  • Session recordings & direct mentor access
  • MCP, multi-agent, and RAG-vs-long-context deep dives
Enroll: Chat on WhatsApp

Frequently asked

Do I need prior AI or ML experience?
No. You need prior coding experience and a basic understanding of software systems. We start from LLM terminology and build up. This is applied AI engineering, not a machine-learning theory or model-training course.
I work full-time. Can I keep up?
Yes, that's why it's weekend-only (Sat & Sun, 9:00–11:30 AM IST). Sessions are recorded, so you can catch up if you miss one. Expect a few hours of optional practice between weekends to finish the labs.
What stack and tools will I use?
Python throughout, with LangChain and ChromaDB for RAG, Neo4j for GraphRAG, Pydantic AI and MCP for agents, LangGraph for multi-agent orchestration, and tools like Langfuse or LangSmith for evals and tracing. You'll also compare frontier APIs against open-weight models.
Do I need an expensive GPU or big API budget?
No. Labs are designed to run on free tiers and small models where possible, and we cover the cost/latency tradeoffs of providers directly so you make budget-aware choices.
Is it live or recorded?
100% live and interactive, you build alongside the instructor and get real-time feedback. Recordings are provided for revision and any missed session.
How do I enroll?
Tap any "Enroll" button to message us on WhatsApp. We'll confirm your seat, share the fee and payment details, and send onboarding instructions.

Eight weeks from now…

…you could be the engineer who ships a real RAG-and-agents system with evals and guardrails, while everyone else is still stuck at the tutorial demo. Reserve your seat for the next cohort.