Now Enrolling

Three Courses.
One Complete Journey.

From mastering prompts to building autonomous agents to training your own language model — everything a modern AI developer needs, taught code-first.

Level 1 — Beginner to Advanced

Prompt Engineering

Master the art and science of communicating with AI models. Learn to craft prompts that extract exactly the outputs you need — reliably, consistently, and at scale. The essential skill for anyone building with AI.

⏱ 18 hours total
📋 18 sessions
🔬 18 hands-on labs
📜 Certificate included
🎥 Recordings included
₹20,000/learner
Inclusive of all taxes
Enroll Now →
Full Curriculum — 6 Modules · 18 Sessions
1

Foundations of Prompting

Sessions 1–3 · 3 hrs
+
S1
How LLMs Work — tokens, context windows, temperature, sampling
Lab: Explore how parameter changes alter outputs on the same prompt
S2
Zero-shot & Few-shot prompting — examples, format, role priming
Lab: Build a few-shot classifier for customer support tickets
S3
Prompt anatomy — instructions, context, examples, output format
Lab: Dissect and improve 5 broken production prompts
2

Reasoning Techniques

Sessions 4–6 · 3 hrs
+
S4
Chain-of-Thought (CoT) — step-by-step reasoning, zero-shot CoT
Lab: Solve math & logic puzzles — CoT vs direct answer comparison
S5
Tree of Thoughts & Self-Consistency — multi-path reasoning
Lab: Build a ToT prompt that outperforms vanilla CoT on ambiguous tasks
S6
ReAct prompting — reason + act cycles for tool-augmented tasks
Lab: ReAct agent that searches Wikipedia to answer factual questions
3

Structured Output & JSON Mode

Sessions 7–9 · 3 hrs
+
S7
JSON mode, response_format, Pydantic schema enforcement
Lab: Invoice extractor — structured output from unstructured emails
S8
XML, Markdown, and custom structured formats for downstream parsing
Lab: Build a prompt pipeline that feeds structured output into a database
S9
Function calling as structured prompting — OpenAI, Claude, Gemini
Lab: Multi-function calling agent with parallel tool execution
4

Advanced Prompt Patterns

Sessions 10–12 · 3 hrs
+
S10
Meta-prompting — prompts that write prompts; automatic optimization
Lab: Auto-generate and A/B test 10 variants of a customer service prompt
S11
System prompt architecture — personas, constraints, grounding strategies
Lab: Build a constrained customer-facing AI persona that never breaks character
S12
Multi-turn conversation management — context windows, compression, steering
Lab: Chat application with rolling summary to handle long conversations
5

Safety & Robustness

Sessions 13–15 · 3 hrs
+
S13
Prompt injection attacks — taxonomy, real examples, detection patterns
Lab: Red-team your own prompt — find and patch 3 injection vectors
S14
Input/output validation — guards, classifiers, moderation APIs
Lab: Build a safe AI assistant with input sanitisation and output filtering
S15
Jailbreaks, bias, hallucination — root causes and mitigation
Lab: Hallucination detector that verifies model claims against a ground truth
6

Production Prompting & Capstone

Sessions 16–18 · 3 hrs
+
S16
Prompt versioning, A/B testing, and evaluation frameworks
Lab: Set up PromptLayer or Langfuse to track prompt performance over time
S17
Cost optimisation — token counting, caching, batching, model selection
Lab: Reduce a 3,000-token prompt by 40% with zero quality loss
S18
🎓 Capstone: End-to-end prompt pipeline — extraction, reasoning, safe output
Lab: Build and deploy a complete document-processing pipeline with evaluation

Who This Is For

  • Developers new to working with LLMs
  • Product managers using AI tools
  • Engineers adding AI features to apps
  • Anyone who works with GPT, Claude, or Gemini daily

Prerequisites

  • Basic Python (variables, functions, loops)
  • Familiarity with LLM APIs (helpful, not required)
  • No ML math or research background needed

What's Included

  • 18 × 1-hour live sessions
  • Session recordings
  • Full GitHub repo with all lab code
  • Certificate of completion
  • 30-day post-course Q&A support
Level 2 — Intermediate · Most Popular

Building AI Agents

Go from curious developer to AI agent architect. Learn to build, orchestrate, and deploy production-grade autonomous agents using LangGraph, CrewAI, OpenAI Agents SDK, and more. The most in-demand AI skill in 2025.

⏱ 18 hours total
📋 18 sessions
🔬 18 hands-on labs
🚀 Deploy in capstone
📜 Certificate included
₹25,000/learner
Inclusive of all taxes
Enroll Now →
Full Curriculum — 6 Modules · 18 Sessions
1

Foundations

Sessions 1–3 · 3 hrs
+
S1
What Is an AI Agent? — capability spectrum, the agent loop, 4 design patterns
Lab: Build raw ReAct agent (~50 lines, no framework)
S2
Agent Architectures — ReAct, Plan-and-Execute, Reflexion, Tree of Thoughts
Lab: Run same research task in ReAct vs Plan-and-Execute
S3
Prompt Engineering for Agents — system prompts, CoT, structured output, failure modes
Lab: Fix 3 broken agent prompts (no code changes allowed)
2

Tool Use & Function Calling

Sessions 4–5 · 2 hrs
+
S4
Core Mechanics — tool loop, OpenAI / Claude / Gemini formats, parallel calls, MCP
Lab: Multi-tool agent (web search, calculator, file reader, clock)
S5
Advanced Patterns — tool chaining, code execution, database tools, smolagents CodeAgent
Lab: Build a minimal MCP server exposing 2 tools → connect to Claude
3

Memory Systems & RAG

Sessions 6–7 · 2 hrs
+
S6
Memory Architecture — 4 types, Chroma/Pinecone/pgvector, rolling summaries
Lab: Customer support agent with persistent Chroma memory
S7
RAG as Agent Memory — Agentic RAG, Self-RAG, CRAG, reranking (Cohere Rerank)
Lab: Agentic RAG pipeline over a 50-page PDF with CRAG fallback
4

Frameworks Deep Dive

Sessions 8–11 · 4 hrs
+
S8
LangGraph — stateful graphs, nodes, edges, checkpointing, human-in-the-loop
Lab: Research agent that pauses for approval before each web search
S9
CrewAI — Agent, Task, Crew; sequential & hierarchical processes; CrewAI Flows
Lab: 3-agent content crew: Researcher → Writer → Editor
S10
OpenAI Agents SDK + AutoGen — handoffs, guardrails, tracing, GroupChat
Lab: Customer support triage with billing, technical & general agents
S11
smolagents & Framework Comparison — LangGraph vs CrewAI vs OpenAI SDK vs smolagents
Lab: Research + summarise pipeline in 3 frameworks side-by-side
5

Multi-Agent Orchestration

Sessions 12–13 · 2 hrs
+
S12
Multi-Agent Patterns — supervisor, swarm/handoffs, hierarchical, parallel fan-out
Lab: Hierarchical system — supervisor + web-researcher + data-analyst + writer
S13
Agent Communication — MCP vs A2A, shared state, resource management, debugging traces
Lab: 2 debate agents + judge agent; visualise trace in LangSmith
6

Evaluation, Safety & Production

Sessions 14–18 · 5 hrs
+
S14
Agent Evaluation — LLM-as-judge, trajectory eval, tool unit tests, Langfuse
Lab: 10-task eval dataset; identify lowest-scoring tool & fix it
S15
Safety & Guardrails — prompt injection, tool misuse, cost caps, HIL gates
Lab: Add injection detector + PII redactor + confirm-before-email to agent
S16
Production Architecture — async/await, stateless vs stateful, Modal / FastAPI deployment
Lab: Deploy LangGraph agent to Modal with Postgres checkpointing
S17
Observability & LLMOps — instrument every call, Langfuse dashboard, KPIs
Lab: Instrumented dashboard — identify slowest tool & highest-cost task
S18
🎓 Capstone: Full Autonomous Research Assistant on Modal
Lab: LangGraph + RAG + code execution + guardrail + REST endpoint

Who This Is For

  • Python developers ready to ship AI products
  • Engineers adding AI to existing systems
  • Tech students building portfolio projects
  • Teams seeking corporate AI training

Prerequisites

  • Python proficiency (functions, classes, async)
  • Basic LLM / API familiarity
  • No ML math required

What's Included

  • 18 × 1-hour live sessions
  • Session recordings
  • Full GitHub repo access
  • Certificate of completion
  • 30-day post-course support
Level 3 — Advanced · Deep Dive

LLMs from Scratch

Stop using LLMs as black boxes. Build, train, and fine-tune a transformer language model from first principles using pure PyTorch. Understand exactly how GPT-style models work — from tokenization to RLHF.

⏱ 18 hours total
📋 18 sessions
🔬 18 hands-on labs
🏗️ Build a GPT from scratch
📜 Certificate included
₹28,000/learner
Inclusive of all taxes
Enroll Now →
Full Curriculum — 6 Modules · 18 Sessions
1

Foundations: Text as Data

Sessions 1–3 · 3 hrs
+
S1
Tokenization — BPE, WordPiece, SentencePiece; vocab sizes; encoding trade-offs
Lab: Implement BPE from scratch; compare against tiktoken on real text
S2
Embeddings — word2vec intuition, positional encoding, learned vs sinusoidal
Lab: Visualise embedding space; probe for semantic & syntactic structure
S3
Language Modelling Objective — next-token prediction, cross-entropy, perplexity
Lab: Train a tiny bigram model; measure perplexity on held-out text
2

Transformer Architecture

Sessions 4–7 · 4 hrs
+
S4
Self-Attention from scratch — Q, K, V matrices; scaled dot-product; masking
Lab: Implement single-head attention in NumPy; visualise attention weights
S5
Multi-Head Attention & Feed-Forward layers — why multiple heads work
Lab: Add multi-head attention; compare attention patterns on different heads
S6
Residual connections, Layer Norm, Dropout — training stability deep dive
Lab: Ablation study — remove residuals or LayerNorm; measure training collapse
S7
Full GPT-2 architecture — decoder-only, causal masking, weight tying
Lab: Assemble complete GPT-2 architecture in PyTorch from scratch (~300 lines)
3

Pre-Training

Sessions 8–10 · 3 hrs
+
S8
Training loop — AdamW, learning rate schedules, gradient clipping, mixed precision
Lab: Pre-train a 10M-param GPT on TinyShakespeare; sample text
S9
Scaling laws — model size, data, compute — Chinchilla & implications
Lab: Run scaling experiments; plot loss vs model size curves
S10
Data pipelines — dataset curation, deduplication, tokenised sharding at scale
Lab: Build a streaming dataloader for a multi-file dataset
4

Fine-Tuning & Alignment

Sessions 11–13 · 3 hrs
+
S11
Supervised Fine-Tuning (SFT) — instruction tuning, chat templates, Alpaca format
Lab: Fine-tune GPT-2 on a QA dataset; compare zero-shot vs SFT outputs
S12
LoRA & QLoRA — parameter-efficient fine-tuning theory and practice
Lab: LoRA fine-tune Llama 3.2 1B on a custom task in under 30 min
S13
RLHF & DPO — reward modelling, PPO intuition, direct preference optimisation
Lab: DPO training on a small preference dataset; evaluate helpfulness shift
5

Inference & Optimisation

Sessions 14–15 · 2 hrs
+
S14
Quantisation — INT8, INT4, GPTQ, AWQ; accuracy vs speed trade-offs
Lab: Quantise Llama model; benchmark latency & memory across 4 formats
S15
KV cache, speculative decoding, Flash Attention, continuous batching
Lab: Measure TPS before/after Flash Attention on your GPT
6

Evaluation, Safety & Capstone

Sessions 16–18 · 3 hrs
+
S16
LLM Evaluation — MMLU, HellaSwag, TruthfulQA; LLM-as-judge; custom evals
Lab: Run lm-evaluation-harness on your fine-tuned model; interpret results
S17
Safety & Alignment — harmful output categories, red-teaming, Constitutional AI
Lab: Red-team your model; add simple Constitutional AI filter
S18
🎓 Capstone: Fine-tuned model + LoRA adapter served via FastAPI
Lab: Full pipeline — pre-trained base → LoRA fine-tune → quantise → REST API

Who This Is For

  • ML engineers wanting deep understanding
  • Researchers moving into applied AI
  • Developers who want to fine-tune models
  • Teams that need custom LLMs

Prerequisites

  • Python & PyTorch basics
  • Linear algebra & calculus fundamentals
  • Completing Prompt Engineering course (recommended)

What's Included

  • 18 × 1-hour live sessions
  • Session recordings
  • Full GitHub repo access
  • Pre-configured Colab notebooks
  • Certificate of completion
  • 30-day post-course support
Pricing

Simple, Transparent Pricing

Choose a single course or save with the complete bundle. Corporate packages available — contact us for a custom quote.

Prompt Engineering

Level 1 · Beginner to Advanced

₹20,000
per learner · all taxes included
  • 18 live sessions
  • Session recordings
  • GitHub repo access
  • Certificate of completion
  • 30-day support
Enroll Now →
Most Popular

Building AI Agents

Level 2 · Intermediate

₹25,000
per learner · all taxes included
  • 18 live sessions
  • Session recordings
  • GitHub repo access
  • Certificate of completion
  • 30-day support
Enroll Now →

LLMs from Scratch

Level 3 · Advanced

₹28,000
per learner · all taxes included
  • 18 live sessions
  • Session recordings
  • GitHub repo + Colab notebooks
  • Certificate of completion
  • 30-day support
Enroll Now →
Best Value

Complete Bundle

All 3 Courses · Full AI Journey

Save ₹13,000
₹73,000
₹60,000
all 3 courses · all taxes included
  • All 3 courses (54 sessions)
  • All recordings & repos
  • 3 certificates
  • Priority support
  • 90-day post-course access
Get Bundle →
FAQ

Common Questions

Everything you need to know before enrolling. Don't see your question? Contact us.

Are the sessions live or pre-recorded? +
All sessions are live and interactive. However, every session is recorded and made available to enrolled students within 24 hours, so you can catch up if you miss a session or want to revisit the material.
What if I'm not satisfied with the course? +
We offer a satisfaction guarantee — if you attend the first 3 sessions and feel the course isn't right for you, we'll process a full refund, no questions asked.
Do you offer corporate or team discounts? +
Yes. For teams of 5 or more, we offer customised corporate packages with dedicated instructors, custom curriculum, flexible scheduling, and volume discounts. Contact us for a tailored quote.
What are the technical requirements? +
A laptop with Python 3.10+ and a code editor (VS Code recommended). For the LLMs from Scratch course, access to a GPU (Google Colab Pro works fine — we provide pre-configured notebooks). All other courses only need a free API key.
Can I take courses in any order? +
Prompt Engineering → Building AI Agents → LLMs from Scratch is the recommended path, but each course can be taken independently. The Agents course assumes Python and basic LLM familiarity. The LLMs course assumes Python and some ML background.
When is the next cohort? +
New cohorts start monthly. Contact us to find out the exact next cohort dates and availability. We keep cohorts intentionally small so spots fill up — reach out early to secure yours.

Start Your AI Journey Today

Join the next cohort and go from where you are now to deploying real AI systems — in as little as 18 hours.

✦ Enroll Now Learn About Us →