Cohort 2 · Starts Sunday, 24 May, 2026
From Software Engineer to AI Engineer in 75 Days.
Live cohort-based bootcamp that takes you from writing code to shipping production-grade AI systems — RAG, agents, multi-agent orchestration, fine-tuning, and deployment. Taught by practitioners. Built with rigor.
22
Live sessions
8
Projects + Capstone
500
Seats (intentional)
Is this for you?
Three kinds of engineers get the most out of this.
Pick the one closest to you. We'll show you how the bootcamp translates into your reality.
The stack you'll own
Every tool real AI engineers use in 2026.
Not a "we'll mention these" list. You'll build, break, and ship with each of these before Day 75.
Python
Core
Groq
Fast Inference
Qdrant
Vector DB
FastAPI
Backend
HuggingFace
Models + Data
semantic-router
Routing
LangChain
LLM Framework
LangGraph
Orchestration
LangSmith
Observability
RAGAS
Eval Framework
Docling
Doc Parsing
AWS
Cloud Deploy
ElevenLabs
Voice TTS
Gemini
Frontier LLM
OpenAI
Frontier LLM
Claude
Frontier LLM
AntiGravity
Agent Framework
Unsloth
Fine-tuning
Ollama
Local SLMs
vLLM
Model Serving
MCP
Tool Protocol
The curriculum
Five arcs. Twenty-two sessions. One shipped portfolio.
Each arc ends with a real project you'll submit, review, and keep. Click an arc to see what's inside.
S01
Kickoff: Welcome to AI Engineering 2.0
Dhaval · Siddhant · Guest AI Engineer
- Welcome & Team Introductions
- Cohort roadmap & Learning flow
- Development Setup: GitHub, Colab, AntiGravity
- AI Engineering 2026 Market Trends
S02
AI Landscape, Python & AI Stack
Dhaval · Siddhant
- AI landscape
- Python fundamentals & OOP
- REST APIs, FastAPI & LLM Integration
- Layers of AI Engineering stack
S03
LLMs, Embeddings & VectorDBs
Dhaval · Siddhant
- Large Language Models: attention, tokens & context windows
- Embeddings & Semantic similarity
- Vector databases with Qdrant
- Model economics & Selection
S04
RAG, LangChain & Evals Intro
Dhaval · Siddhant
- RAG pipeline & Document Ingestion
- LangChain Basics & LCEL
- Hybrid Search: Improving Retrieval
- Evaluation Basics: RAGAS, faithfulness, answer relevance, context precision
Project 1: RAG Q&A System
Build a RAG Q&A system over any document set of your choice.
Set up LangSmith tracing and run RAGAS evals.
Submit your eval scores (faithfulness + relevance) + LangSmith trace link.
S05
Agentic AI Foundation
Dhaval · Siddhant
- What is ReAct?
- Tool calling & reasoning loop
- Memory systems for agents
- Reliability caveats in production
S06
Agent Orchestration & LangGraph
Dhaval · Siddhant
- Workflow state machines
- Conditional branches & loops
- Build your first Agent in Langchain
- Routing, guardrails & compliance (HIPAA, GDPR)
- Trajectory evaluations
Project 2: Multi-tool Agent
Build a multi-tool agent with 3+ tools using Claude function calling.
Implement it in LangGraph as a state machine (not just a chain).
Add a trajectory eval, measure whether your agent took the right path,
not just whether it reached the right answer.
Submit: agent code + trajectory eval results + LangSmith trace.
S07
LangSmith & AWS Deployment
Dhaval · Siddhant
- Traces, runs & structured feedback
- Eval datasets & monitoring
- Debug deployed AI systems
- Deploy apps to AWS
S08
Evals & Guardrails in Production
Dhaval · Siddhant
- Offline / online eval systems
- Human-in-the-loop review
- Detecting performance drift
- Beyond RAG agent evals
Project 3: Evaluation Harness
Add a complete eval harness to your Project 1 RAG system or Project 2 agent.
Required: offline eval set (min 20 examples) + LLM-as-judge (Claude) + at least one guardrail.
Run automated red-teaming and document what you found.
Submit: eval harness code + results report + red-team findings.
S09
Model Context Protocol (MCP)
Dhaval · Siddhant
- MCP : Anthropic's Model Context Protocol
- How to build your own MCP servers?
- Connect external MCPs with your agents
S10
Context Engineering
Dhaval · Siddhant
- Importance of Context in complex AI Systems
- Context Engineering Fundamentals
- Improving Context Propagation in your agents
- Introduction to deepagents
S11
Multi-Agent Systems
Dhaval · Siddhant
- Multi-agent architectures
- Subgraphs & parallel execution
- Circuit breakers & fallbacks
- Workflow vs Agentic Architecture Design
Project 4: Multi-Agent System
Build a Multiagent AI System: LangGraph multi-agent + MCP server + RAG.
Your system must have: 2+ agents with defined roles, at least 1 MCP tool, persistent memory via LangGraph checkpointer, and LangSmith tracing throughout.
Submit: architecture diagram + code + LangSmith dashboard screenshot.
S12
Multimodal AI Engineering
Dhaval · Siddhant
- Understanding Multimodal AI
- Vision Transformers Architecture
- Building a mutimodal pipeline with Gemini
- Multimodal RAG: index and retrieve images alongside text
S13
Personal Branding + Text-2-SQL
Hemanand · Siddhant
- Position yourself as an AI Engineer.
- GitHub portfolio storytelling
- Levereging LinkedIn & Twitter to build your brand.
- Text-2-SQL architecture
- Self-correcting SQL agents
S14
SLMs & Local AI
Dhaval · Siddhant
- What are SLMs?
- Casestudies of SLMs in Production
- Popular SLMs & their forte
- Ollama & vLLM: Running Models Locally
- Build local AI API endpoint
Project 5: Local LLM API
Deploy a local SLM (Qwen 2.5-3B) as a production REST API using vLLM.
Build a simple router: classify each query and send simple ones to the SLM, complex ones to Claude Sonnet.
Measure and report: latency comparison + cost savings at 10K requests/day.
Submit: router code + cost analysis + latency benchmarks.
S15
Deep Work + Industry Case Study
Hemanand · Guest
- Protect focused time
- Reduce context switching
- Production AI failures & lessons
- Architecture tradeoff analysis
S16
Voice AI
Dhaval · Siddhant
- Basics of real-time voice systems
- Low-latency voice systems
- Handling interruptions and latency
- Multi-turn context recovery
- Building a voice agent
Project 6: Voice Agent
Build a voice agent OR a multimodal agent
Voice track: agent must respond in under 800ms end-to-end. Test with 10 different queries.
Multimodal track: agent must extract structured data from 3 different document types.
Submit: working demo recording + latency/accuracy metrics.
S17
Advanced AI Engineering
Dhaval · Siddhant
- Cost optimization at scale
- Semantic caching
- Smart model routing
- Advanced error handling, retries & rate-limiting
S18
Fine-Tuning & Synthetic Data
Dhaval · Siddhant
- Fine-tune vs RAG: When to opt for what?
- LoRA & QLoRA training
- Synthetic dataset pipelines
- Data filtering & deduplication
S19
AI Product Thinking + Reinforcement Learning
Dhaval · Siddhant · Karan
- Turning AI capabilities into usable, valuable products
- Thinking like an AI product builder, not just a model user
- Core intuition behind reinforcement learning in real-world AI
- Reward design: designing rewards for better behavior
Project 7: Finetune Your Custom Model
Generate a synthetic dataset
Fine-tune Qwen 2.5-1.5B with LoRA on your dataset (Unsloth + HuggingFace TRL).
Submit: dataset + training config + before/after eval scores + 1-page findings report.
S20
AI Security
Dhaval · Siddhant
- OWASP Top 10 for AI
- Prompt injection attacks
- Hallucination drift risks
- Context poisoning defense
- Sandboxed tool execution
- Defending your system from adversarial attacks
S21
Capstone Project
Dhaval · Siddhant
- Build complete AI product
- End-to-end system showcase
- Production-ready architecture
- Final polish & review
🎓 Capstone Project
Choose one of the capstone project & build it end-to-end based on the concepts learned throughout the cohort.
Create product demos & Present in showcase.
S22
Interview Preparation + Capstone Project Showcase
Dhaval · Siddhant · Hemanand · Guest Panel
- Present capstone to panel
- AI system design interviews
- Portfolio review feedback
- Live demo & critique
The projects
Eight shipped pieces. Each one goes on your GitHub.
Not toy assignments. Real systems with real evals, real tracing, real submissions.
Session 4 · End of Arc 01
Build a production RAG Q&A system
Pick any document set you care about. Build the full RAG pipeline: ingestion, chunking, embeddings, retrieval, and generation. Experiment with chunking strategies to improve ingestion & retrieval.
Tech you'll use
LangChain
Qdrant
Groq
Docling
What you'll submit
→ Code for end-to-end RAG Pipeline
→ Architecture diagram
→ Set-up guidelines & README.md
→ Architecture diagram
→ Set-up guidelines & README.md
Session 6 · Arc 02
Multi-Tool Agent with Trajectory Eval
Build an agent that uses at least 3 tools via function calling. Build a ReAct agent using LangChain. Then measure whether it takes the right path, not just whether it reaches the right answer.
Tech you'll use
LangChain
Groq
Function Calling
Python
What you'll submit
→ Agent code (3+ tools)
→ Trajectory eval results
→ Set-up guidelines & README.md
→ Trajectory eval results
→ Set-up guidelines & README.md
Session 8 · End of Arc 02
Add a real eval harness to your system
Take your Project 1 RAG or Project 2 agent and give it the evaluation infrastructure that production systems actually need. Offline eval set, LLM-as-judge, guardrails, and automated red-teaming.
Tech you'll use
RAGAS
LangSmith
Claude as judge
Red-team tooling
What you'll submit
→ Offline eval set (20+ examples)
→ LLM-as-judge evaluator code
→ At least 1 working guardrail
→ Red-team findings report
→ LLM-as-judge evaluator code
→ At least 1 working guardrail
→ Red-team findings report
Session 11 · Arc 03
A Multiagent System with MCP, RAG & Memory
Put it all together. Multi-agent LangGraph with at least two roles. One MCP server you built. RAG grounding. Persistent memory via checkpointer. Full observability via LangSmith.
Tech you'll use
LangGraph
MCP
Qdrant
LangSmith
Claude
What you'll submit
→ System architecture diagram
→ Full code repo
→ LangSmith dashboard screenshot
→ 2+ agents with defined roles, 1 MCP tool, persistent memory
→ Full code repo
→ LangSmith dashboard screenshot
→ 2+ agents with defined roles, 1 MCP tool, persistent memory
Session 14 · Arc 04
Ship a local SLM with a smart router
Deploy Qwen 2.5-3B as a production REST API using vLLM. Build a router that classifies each query — simple ones go to your local SLM, complex ones to Claude Sonnet. Measure and prove the cost savings.
Tech you'll use
vLLM
Qwen 2.5
FastAPI
Claude
What you'll submit
→ Router code + deployment config
→ Latency comparison (SLM vs Claude)
→ Cost analysis at 10K requests/day
→ Benchmark results
→ Latency comparison (SLM vs Claude)
→ Cost analysis at 10K requests/day
→ Benchmark results
Session 16 · Arc 04
Voice agent or multimodal agent — your choice
Voice track: Build an agent that responds end-to-end in under 800ms across 10 real queries. Handle interruptions and silences.
Multimodal track: Build an agent that extracts structured data from 3 different document types with high accuracy.
Tech you'll use
Deepgram
ElevenLabs
LiveKit
Gemini (multimodal)
What you'll submit
→ Working demo recording
→ Latency or accuracy metrics
→ Code repo
→ Edge case handling writeup
→ Latency or accuracy metrics
→ Code repo
→ Edge case handling writeup
Session 19 · Arc 05
Fine-tune your own SLM from scratch
Generate a synthetic dataset for a narrow task. Curate it — dedupe, quality filter, diversity sample. Fine-tune Qwen 2.5-1.5B with LoRA using Unsloth and HuggingFace TRL. Show the before/after eval scores.
Tech you'll use
Unsloth
HuggingFace TRL
LoRA / QLoRA
Qwen 2.5
What you'll submit
→ Synthetic dataset (curated)
→ Training config + run logs
→ Before/after eval scores
→ 1-page findings report
→ Training config + run logs
→ Before/after eval scores
→ 1-page findings report
Sessions 21–22 · Final
Capstone — Production-ready AI system
A complete system that demonstrates the full cohort arc. Must include: RAG or multi-agent orchestration, an eval harness, SLM + frontier model routing, a security audit, and a real deployment. Judged by external AI engineers on production readiness.
What the panel evaluates
Architecture quality
Eval rigor
Cost model
Security
Shipability
What you'll present
→ Live demo
→ Architecture walkthrough
→ Eval results
→ Cost model & unit economics
→ Security audit findings
→ Architecture walkthrough
→ Eval results
→ Cost model & unit economics
→ Security audit findings
Why this bootcamp?
The honest comparison.
Other paths can work. Here's exactly what you get from each.
Self-study
Generic GenAI course
This bootcamp
Live, practitioner-led sessions
✕ None
~ Recorded
✓ 22 sessions
Production eval harnesses
✕ Rare
✕ Glossed over
✓ Dedicated arc
Multi-agent + MCP in-depth
~ Scattered docs
✕ Usually missing
✓ Full session each
Hands-on fine-tuning (LoRA)
✕ Rare
~ Theory only
✓ Your own SLM
Shipped projects on GitHub
~ On your own
~ 1–2 toys
✓ 7 + Capstone
External panel review of your work
✕ None
✕ None
✓ Capstone panel
Full refund if not the right fit
✕ N/A
~ Fine print
✓ Before 31 May, 2026
Who you'll learn from:
Three practitioners. One split: Build. Orchestrate. Distribute.
Each teaches what they actually do day-to-day.
Dhaval Patel
Lead Instructor · Build
Ex-NVIDIA. 17+ years in AI and data, 1.5M+ subscribers on YouTube teaching AI & ML. Core architect of the AI Engineering curriculum.
Teaches
LLM fundamentals · RAG · Multi-agent systems · AI system design · Capstone project
Hemanand Vadivel
Co-Instructor · Distribute
Ex-Edgewell. Builds the bridge from engineering skill to hireable presence. Leads the career & craft side of the cohort.
Teaches
Productivity systems · Stakeholder management · LinkedIn strategy · Personal branding
Siddhant Pandey
Co-Instructor · Orchestrate
AI Research Engineer. Runs the live labs and production code walkthroughs across every session. Your real-time debugging partner.
Teaches
Hands-on coding · Live debugging · Eval design · Observability · Production patterns
Investment
The earlier you join, the less you pay.
One cohort. One bootcamp. Two pricing windows.
★ Inner Circle
Join until Sunday, 17 May, 2026
US$630
→
US$840
Save US$210 + get access to a curriculum-enhancing session
Standard
Join from Monday, 18 May, 2026
US$840
One-time payment
Everything included in both
75-day intensive, 22 sessions
Live weekend cohorts with Q&A
7 projects + Capstone on GitHub
External panel review of Capstone
1 year access to recordings
Certificate of completion
Gen AI & DS Bootcamp (US$291 value)
Refund window - before 31 May, 2026
Enroll now →
Cohort 2 begins Sunday, 24 May, 2026.
Questions?
Everything you'd reasonably want to know.
Filter by category, or just scroll.
No. If you have at least 2 years of software engineering or coding experience, you're ready. We cover AI from an engineering lens.
The Inner Circle is early enrollment for Cohort 2, open until May 10, 2026. Inner Circle members get a dedicated live session a few days before the full cohort launches around May 24th to help enhance the curriculum through their feedback. You’re not just enrolling early, you’re influencing what gets built.
Yes. Every enrollment includes full access to the Gen AI & Data Science Bootcamp at no extra cost. Think of it as your reference library throughout the journey.
Saturdays and Sundays, 4–7 PM IST. Sessions are fully live and interactive, with a strong focus on hands-on practice, real-time problem solving, and Q&A. Recordings are available for revision.
Around May 24th, 2026, few days after the Inner Circle session.
All live sessions are recorded and uploaded within 24–48 hours. You can watch them at your own pace.
No. You keep access to all recordings for 1 year from your enrollment date.
You're securing your seat now. Full bootcamp access opens when Cohort 2 officially launches around May 24th, 2026.
We strongly advise against it. This bootcamp moves fast and assumes you can already write code confidently. If you're early in your career, start with the Gen AI & Data Science Bootcamp first. Build the foundation, then come back.
Software engineers moving into AI roles, developers adding AI to existing products, and tech leads or architects designing AI systems. If you write code for a living and want to build real AI systems, this is for you.
You can ask questions anytime on Discord. During live sessions, instructors solve problems in real time. You're never stuck alone.
The AI Engineering Bootcamp focuses on building your skills, portfolio, and online presence. The Gen AI & DS Bootcamp (included with your enrollment) has dedicated job assistance: resume builder, LinkedIn optimization, portfolio website, and mock interviews.
Inner Circle enrollment is open until May 10, 2026. After that date, the price goes up when the full cohort launches. This is the only window to lock in this price.
The amount you paid for the Gen AI & DS Bootcamp is fully adjusted and deducted from your AI Engineering Bootcamp price.
We don't offer a standalone version. The Gen AI & DS Bootcamp is intentionally bundled as a foundational reference throughout the program. If you already own it, you're eligible for a reduced price and your earlier purchase is not wasted.
Full refund, no questions asked, if you request by 31st May,2026. After that, no refunds are available.
Full refund if requested by 31st may 2026. After that date, no refunds are available. Check the complete policy: https://codebasics.io/refund-policy
No. Once your existing purchase is applied as a subsidy to reduce your price, that original purchase becomes non-refundable.
You get back the amount you actually paid for the AI Engineering Bootcamp. Your original purchase stays intact and you keep access to it.
Recommended:
• OS: Windows 11
• Processor: Intel Core i7 (10th Gen+) or AMD Ryzen 7 (4th Gen+). An i5 works if you're not focused on local model training.
• RAM: 8GB minimum, 16GB recommended
• Storage: 512GB SSD strongly recommended
• GPU: NVIDIA GTX 1660 or higher for deep learning and GPU-accelerated tasks
This covers all bootcamp work comfortably. You'd only need stronger hardware if you plan to fine-tune small LLMs locally.
• OS: Windows 11
• Processor: Intel Core i7 (10th Gen+) or AMD Ryzen 7 (4th Gen+). An i5 works if you're not focused on local model training.
• RAM: 8GB minimum, 16GB recommended
• Storage: 512GB SSD strongly recommended
• GPU: NVIDIA GTX 1660 or higher for deep learning and GPU-accelerated tasks
This covers all bootcamp work comfortably. You'd only need stronger hardware if you plan to fine-tune small LLMs locally.
This is your curriculum enhancement session. All Inner Circle members come together live with the team to share feedback and help enhance the Cohort curriculum.
The session will be scheduled a few days before the cohort launch on 24th May 2026. The exact date will be communicated in advance. What you share here directly shapes what gets built.
The session will be scheduled a few days before the cohort launch on 24th May 2026. The exact date will be communicated in advance. What you share here directly shapes what gets built.
The best time was yesterday.
The second best is now.
500 seats · Inner Circle closes 17 May, 2026 on request
Enroll now →
Cohort 2 · Starts Sunday, 24 May, 2026.
Log in with Google
Log in with Linkedin