Batch 1 (Nov 2026 Intake) — Starts Mon, 30 Nov 2026

What you'll learn

1. 1. Week 1: GAIL Modules 1–2: Modern LLM Architecture, Advanced Prompt Engineering & EvalAI vs ML vs Deep Learning vs Generative AI; why foundation models changed software engineering · How LLMs work: tokens, embeddings, context windows, pre-training, RLHF, multimodal models · Data lifecycle, structured vs unstructured data, labeled vs unlabeled data · ML training paradigms: supervised, unsupervised, reinforcement learning · Hallucinations, latency, token economics, and when NOT to use an LLM · Google's foundation model landscape: Gemini (Nano, Flash, Pro, Ultra), Gemma open weights, Imagen, Veo, Chirp, Lyria · Model selection trade-offs: context size, cost, latency, privacy, edge (LiteRT) vs cloud · Zero-shot, few-shot, role prompting, Chain-of-Thought, ReAct, and prompt chaining · Sampling parameters: temperature, top-p, top-k, max output tokens, safety filters, system instructions · Building a 20-case golden evaluation set to score prompts objectively
2. 2. Week 2: Basic Python for AI, APIs & Structured JSONBasic Python fast-track first (pre-recorded Fast-Start Setup Pack + lab clinic): syntax & indentation, variables and data types, if/else and loops, functions, lists & dictionaries, reading/writing files and JSON, modules, pip and virtual environments (venv), calling a REST API with requests, and your first simple LLM call with the Gemini Python SDK · Python 3.11+ for AI: virtual environments, type hints, requests, environment variables · Calling Gemini & OpenAI-compatible APIs from Python: multi-turn chat, system prompts, streaming tokens · Guaranteed structured outputs using Pydantic schemas and JSON mode · Production resilience: exponential backoff retries, rate-limit handling, token counters, and cost logging · SQL essentials for AI engineers: queries, joins and safe read-only text-to-SQL with Postgres/SQLite · Writing simple pytest unit tests for your functions
3. 3. Week 3: GAIL Modules 3–4 (I): AI Layers, Responsible AI, Gemini Workspace & NotebookLM + MILESTONE 1 DEPLOYED APPThe 5 layers of an enterprise GenAI stack: Infrastructure (TPUs/GPUs), Models, Platform (Agent Platform / Vertex AI), Agents, Applications (Workspace / Gemini Enterprise) · Responsible AI & Google's Secure AI Framework (SAIF): fairness, bias, privacy, explainability, human-in-the-loop · Building & deploying web UIs for AI in Python (Streamlit / Gradio) to a public URL · Google Workspace with Gemini across Gmail, Docs, Sheets, Slides, Meet, and Vids · Building custom Gems for repeatable team workflows · Gemini Enterprise & Gemini Notebook (NotebookLM Enterprise) for multi-document grounding and synthesis · Agent Platform (Vertex AI) vs Google AI Studio: decision matrix for enterprise architects
4. 4. Week 4: Embeddings, Semantic Search & Vector Databases (Theory + Python)How text & multimodal embeddings work: cosine similarity, dot product, dimensionality, and domain drift · Document ingestion pipeline: PDF/Markdown/HTML parsing, fixed vs recursive vs semantic chunking, overlap & metadata tags · Vector databases in practice: ChromaDB (local/persistent) and Postgres pgvector (Supabase) · When semantic search fails on exact keywords/SKUs and how hybrid search (BM25 + dense vectors) fixes it
5. 5. Week 5: End-to-End RAG Architecture: Citations, Re-Ranking & EvalThe complete RAG loop: Retrieve -> Re-rank -> Augment -> Generate with strict source citations · Google Cloud grounding offerings: Agent Search (Vertex AI Search), RAG APIs, Grounding with Google Search · Preventing hallucinations: strict 'Answer ONLY from context or say I don't know' prompts + citation verification · Evaluating RAG systems: context precision/recall, faithfulness, answer relevance, and prompt-injection defense
6. 6. Week 6: FastAPI + Docker + CI/CD + Cloud Deployment & MILESTONE 2 RAG APPTurning Python AI scripts into production REST APIs with FastAPI: Pydantic request/response models, streaming endpoints, CORS, API key auth · Containerizing AI backends with Docker (`Dockerfile`, `.dockerignore`, multi-stage builds) · Deploying to Google Cloud Run / Render / Hugging Face Spaces with environment secrets and persistent vector storage · CI/CD with GitHub Actions: tests, eval checks and Docker image build on every push · Container security basics: non-root images, image scanning, secrets in Secret Manager not in code · Multi-cloud mapping for architects: Vertex AI/Cloud Run ↔ Azure OpenAI/Container Apps ↔ Amazon Bedrock/ECS Fargate
7. 7. Week 7: Customer Engagement Suite, Pre-Built AI APIs, GAIL Module 5 (I) Agents & Tool Calling + GAIL MOCK EXAM 1Google Cloud Customer Engagement Suite: Conversational Agents, Agent Assist, Conversational Insights, CCaaS · Pre-built Google Cloud AI APIs: Document AI, Speech-to-Text, Text-to-Speech, Translation, Vision, Video Intelligence, Natural Language · Selecting pre-built APIs vs RAG vs fine-tuning for enterprise RFPs · What makes an AI Agent: reasoning loop (ReAct), memory, planning, and external tools · Deterministic vs generative vs hybrid agents; extensions, functions, data stores, and plugins · Implementing Function Calling / Tool Use in Python: defining JSON tool schemas, executing Python tools safely, and feeding tool outputs back to the model
8. 8. Week 8: Multi-Agent Systems: LangGraph, Google ADK, MCP & A2AStateful agent workflows and multi-agent orchestration (Router, Researcher, Coder/Analyst, Critic/Verifier) · Building production agent graphs with LangGraph / Google Agent Development Kit (ADK) · Model Context Protocol (MCP): connecting agents to standardized data sources and enterprise tools · Human-in-the-loop approval gates, max-step budgets, and tool-error recovery · Building a custom MCP server and an overview of the Agent2Agent (A2A) protocol
9. 9. Week 9: GAIL Module 5 (II): Enterprise Agent Platform, MLOps, Observability & MILESTONE 3 AGENTGemini Enterprise Agent Platform (Vertex AI): Model Garden, Model Builder, AutoML, Agent Builder, Agent Search · MLOps for GenAI: Model Registry, versioning, evaluation pipelines, drift & performance monitoring, Feature Store · Fine-tuning options (supervised tuning, RLHF, distillation, LoRA) vs RAG: cost and maintenance trade-offs · LLM observability and cost control: tracing (OpenTelemetry/Langfuse), token cost per request, caching and model routing
10. 10. Week 10: Business Strategy, ROI Modelling, SAIF Security, OWASP/DPDP & GAIL MOCK EXAM 2Executive GenAI selection framework: scale, customization, latency, privacy, connectivity, and build-vs-buy TCO · Top-down + bottom-up transformation strategy, change management, and KPI/ROI calculation · Secure AI in depth: Google SAIF, IAM least privilege, VPC Service Controls, Security Command Center, data residency & DLP · OWASP Top 10 for LLM Applications and India's DPDP Act 2023 basics for AI systems
11. 11. Week 11: Dual Capstone Build Sprint + AI Career Prep & GAIL Exam BookingScoping a high-impact Dual Capstone that satisfies both hiring managers and enterprise architecture panels · End-to-end engineering sprint: combining FastAPI, Vector DB/RAG, Multi-Agent tools, UI polish, and automated eval suite · Writing the 6-page Executive GenAI Proposal (business case, Google Cloud architecture diagram, model/RAG choices, SAIF security, and 12-month ROI) · Structuring an AI Engineer / GenAI Leader resume and LinkedIn profile around 3 deployed URLs and quantified eval metrics · GitHub portfolio makeover: pinned repositories, architecture diagrams, 2-minute Loom/video demos, clean READMEs · Mock Technical & Leadership Interviews: LLM/RAG system-design scenarios + GAIL situational business questions · Step-by-step official Google Cloud GAIL Exam registration walkthrough (online-proctored vs Pearson VUE test centre)
12. 12. Week 12: Final Exam Cram Clinic & DUAL CAPSTONE DEMO DAYRapid-fire 60-minute GAIL exam cram: high-frequency exam traps, Google Cloud product naming cheatsheet, time-management strategy for 50-60 questions in 90 minutes · Production launch checklist: logging, error boundaries, README, and demo script · Grand Finale Dual Capstone Demo Day in front of instructors and invited industry guests

Final project demo: Dual Master Capstone: Executive Enterprise GenAI Proposal + Deployed Production RAG & Multi-Agent System. Every graduate ships a two-part flagship capstone: (1) A 6-8 page Executive Architecture & Business Proposal mapped to Google Cloud's GAIL framework (use case, KPI/ROI model, Google Cloud solution layer, model trade-offs, grounding design, SAIF security & Responsible AI rollout), AND (2) A live, publicly deployed full-stack RAG & Multi-Agent application in Python (FastAPI + Vector DB + LangGraph/ADK + Docker/Cloud Run) with a clean GitHub repository, architecture diagram, evaluation scorecard, and 2-minute demo video.