00
AI-Native Weekend (free entry)
Two-day live masterclass that runs the month before the cohort. Optional but recommended. You leave with one shipped demo and a sense of whether this discipline fits you.
SHIP1 public demo
Cohort 1 · June 14, 2026
Six weeks. Forty builders. One discipline.
A live, pod-based cohort taught by engineers who shipped production AI last quarter. You leave with six public ships, your own AI-native system in your own GitHub, a governance binder that passes a real model-risk review, and a capstone you can demo in five minutes without hedging.
// Application close · June 7// Cohort start · June 14// Demo Day · July 26
Section 01
Who this cohort is for
Three audiences sit in the same room. They learn from each other on purpose.
TIER · LAUNCH
The Student
College or early-career. Wants to leapfrog the AI job market. Graduates with a system they built, a capability graph that proves their skill, and a pipeline into real AI jobs.
TIER · SCALE
The Professional
Working engineer, PM, analyst, operator. Wants to lead AI transformation at their company or switch into an AI-native role. Graduates with a system installable at their own org.
TIER · ASSURE
The Senior Architect
Staff+ engineer or architect at a regulated shop. Wants the discipline that survives audit. Graduates with a binder examiners accept and a method they can lead with internally.
Not a fit if: you want a no-code tutorial, you refuse to write code in public, or you want a certificate without shipping.
Section 02
Program structure
Live + async. Pods of 8–12. Mandatory weekly public ship.
Section 03
Weekly curriculum
Each week ends with a public ship. No ship, no graduation.
01
AI Fluency & Context Engineering
System prompts with cached context, structured outputs, tool definitions. The shape of an LLM call that does not break in week six.
SHIPPersonal AI OS + prompt pack
02
RAG & Retrieval Systems
Ingestion, chunking, embeddings, hybrid search, reranking, citation-grounded answers. RAG that survives a real corpus.
SHIPWorking RAG app on your own data
03
Agent Engineering
Multi-step agents in LangGraph or the Claude / OpenAI Agents SDK, wired through MCP to real tools. Agents as state machines, not magic.
SHIPMulti-step agent solving a real task
04
Evaluation, Observability & Trust
MOAT WEEKThe moat week. Eval-driven development, OpenTelemetry tracing, audit trails a compliance officer will accept, hallucination quarantine. Plan more than eight hours this week.
SHIPEval suite + audit trail + dashboard
05
Fine-Tuning, MLOps & Production
When fine-tuning beats prompting, LoRA / QLoRA on Modal or HF, deployment with canary rollout, cost tuning. Production discipline.
SHIPDeployed, monitored, cost-tuned agent
06
Capstone & Demo Day
Scope, build, eval, govern, demo. Five-minute pitch, governance binder, live Q&A from a skeptical room. Graduate the way you would defend in production.
SHIPCapstone + governance binder + pitch
Section 04
What you can do by Week 6
These are not aspirations. They are the bar for graduation.
- 01Engineer a system prompt with cached context, structured outputs, and tool definitions a junior dev cannot produce by vibes alone.
- 02Stand up a production-grade RAG pipeline: ingestion, chunking, embeddings, hybrid search, reranking, citation-grounded answers.
- 03Design and ship a multi-step agent with LangGraph or the Claude/OpenAI Agents SDK, wired through MCP to real tools.
- 04Write an eval suite in Braintrust or Langfuse that catches regressions — golden datasets, LLM-as-judge, hallucination checks — and run it in CI.
- 05Instrument OpenTelemetry tracing, produce an audit trail a compliance officer will accept, and explain the difference between a trace and a log.
- 06Decide with numbers when fine-tuning beats prompting, run a LoRA/QLoRA job on Modal or HF, and deploy with canary rollout.
- 07Produce a governance binder: model card, risk assessment, eval report, prompt register, audit trail. Non-negotiable in regulated shops.
- 08Pitch your work in five minutes to a skeptical room and survive the Q&A.
Section 05
Tool stack you will be fluent in
We teach tool-agnostic patterns, but you ship with real tools.
MODELS & APIS
Anthropic Claude (Opus 4.7, Sonnet 4.6, Haiku 4.5), OpenAI GPT-5.x + o4, Google Gemini 3.1 Pro, open-weights Llama / Qwen via Bedrock + Azure AI Foundry.
DEV ENVIRONMENT
Cursor, Claude Code, VS Code, uv, pnpm, GitHub Actions.
RETRIEVAL
pgvector, Pinecone, Weaviate, LlamaIndex, LangChain, Cohere Rerank v3, Voyage voyage-3.
AGENTS
LangGraph, Claude Agent SDK, OpenAI Agents SDK, CrewAI, AutoGen, Model Context Protocol (MCP).
EVALS & OBSERVABILITY
Braintrust, Langfuse, OpenTelemetry, OpenLLMetry, Ragas.
DEPLOYMENT
Vercel AI SDK, Modal, Replicate, Hugging Face, AWS Bedrock, Azure AI Foundry, Fly.io.
Section 06
Pricing summary
Three tiers. Five scholarship seats per tier. Corporate cohorts on request.
TIER · LAUNCH
₹9,999
Students · 4 wks core
TIER · SCALE
₹49,999
Professionals · 8 wks
TIER · ASSURE
₹1,49,999
Senior architects · 12 wks
Section 07
Frequently asked
Yes. No credit card. No trial. No upsell pressure. If you’re serious about building AI, you’re welcome.
About 8 hours per week. Two live sessions (90 minutes each), one pod working session, plus your weekly ship. Doable while working full time.
Basic Python or JavaScript. Comfort with the terminal. We don’t teach coding from scratch, but if you can write a function, you can do this cohort.
No. We teach the methodology, not the artifacts. You leave with your own codebase, shipped to your own GitHub. That’s the point.
They sell courses. We offer a guild with a methodology. They teach theory. We teach shipping. They give certificates. We give you a capability graph, a pod for life, and a credential that’s earned.
Yes. Corporate Capability Workshops start at ₹20 lakh. Individual Scale and Assure tiers are tax-deductible as L&D in many jurisdictions.
Model card. Risk assessment. Eval report. Prompt register. Audit trail documentation. The artifact enterprises actually ask for before they deploy. Your capstone includes one.
Tire-kickers. Lurkers. People who want a certificate without shipping anything. If you won’t commit ~8 hours a week, please don’t apply.
Sessions run at two times: 10 AM IST and 6 PM PT. Pick either. Pods form by timezone. All sessions are recorded; live attendance earns the pod invitation.
Cohort 1 begins June 14, 2026. Cohort 2 opens August 9. Cohort 3 in late October. We run five to six cohorts per year.
Cohort 1 · 40 seats
Apply by June 7. Cohort starts June 14.
We review every application within 48 hours. Acceptance comes with a payment link and onboarding email.