🏢
Enterprise Knowledge Assistant
Agentic RAG over HR, policy and product documents with access control, citations, evaluation and a Databricks Apps UI.
Course DT-GAI-120 · Live online · New batches every month
Generative AI & Agentic AI Engineering Training with Databricks Mosaic AI
Most GenAI courses stop at notebooks and chatbots. In this 10-week, 120-hour live program you build, evaluate, secure and deploy real agentic AI applications on Databricks, the way enterprises actually run them: LLMs, RAG, LangGraph agents, multi-agent systems, MCP, A2A and LLMOps.
📅 10 weeks · 120 hours
🎥 Live classes + recordings
🧪 9 labs + 1 capstone
🎓 Databricks GenAI Engineer Associate prep
◆ Core + Architect tracks
Course highlights
120 hrs80 live + 40 guided lab
10Weekly lab projects + capstone
40+GenAI tools & frameworks
4.9 ★Google rating · 200+ reviews
1200+Professionals trained
📋 Course at a Glance
10 weeks from LLM basics to production AI agents
Every week is 12 hours: 4 weekday live sessions and 1 weekend lab. Core topics are for everyone; ◆ Architect topics add extra depth for professionals with 5-15+ years of experience.
| Week | Module | Hours |
|---|---|---|
| Week 1 | How Generative AI Thinks LLM Foundations, Reasoning Models & Open Models | 12 |
| Week 2 | Talking to AI Effectively Prompt & Context Engineering | 12 |
| Week 3 | Teaching AI Your Company's Data Embeddings, Vector Search & RAG | 12 |
| Week 4 | Making AI Answers Accurate & Trustworthy Advanced RAG, Knowledge Graphs & Evaluation | 12 |
| Week 5 | Building Your First AI Agents AI Agents, LangGraph & Memory | 12 |
| Week 6 | AI Teams That Work Together Multi-Agent Systems: CrewAI, AutoGen, PydanticAI | 12 |
| Week 7 | Connecting AI to Tools & Other Agents Agent Protocols: MCP, A2A & Agent Skills | 12 |
| Week 8 | AI That Sees, Hears, Browses & Codes Multimodal, Browser, Computer-Use & Coding Agents | 12 |
| Week 9 | Running AI Safely in Production LLMOps, Security, Responsible AI & n8n | 12 |
| Week 10 | Designing AI Systems & Getting Hired GenAI Architecture, Capstone & Career | 12 |
| Total | 120 |
👥 Who this course is for
- Freshers and career changers with Python and SQL
- Data engineers and analysts moving into GenAI
- ML engineers adding agents and LLMOps
- Tech leads and architects (5-15+ yrs) designing GenAI platforms
🛠️ What you will build
- RAG chatbot over your own documents
- SQL data-analyst agent with memory
- Your own MCP server used by Claude
- Browser, voice and multimodal agents
- A secured, monitored production deployment
- A scored capstone for your portfolio
◆ Two Tracks, One Batch
Built for freshers and for experienced professionals (5-15+ yrs)
The Core track is for freshers and career changers. The Architect track is added on top for senior engineers, leads and architects: these are the advanced topics expected in 2026 interviews and projects. Every week, Architect learners also submit a 1-page design note reviewed by the trainer.
| ◆ Architect topic | Why it matters | Where |
|---|---|---|
| Reasoning models, SLMs & edge AI | Choosing between reasoning, fast, local and on-device models is now an architecture decision. | Week 1 |
| DSPy & GenAI in pipelines | Systematic prompt optimisation and LLM-powered ETL are natural next steps for data engineers. | Week 2 |
| Enterprise-scale RAG | Access control, CDC indexing and freshness SLAs separate demos from real systems. | Week 3 |
| Knowledge graphs & eval-driven development | Connected reasoning, and quality gates that block bad releases. | Week 4 |
| Agent memory & Lakebase | Long-running agents need durable state and memory, not just chat history. | Week 5 |
| A2A protocol & Agent Skills | Agents from different teams and vendors must interoperate. | Week 7 |
| Multimodal, voice & coding agents | The fastest-growing agent categories in 2026. | Week 8 |
| AI security, responsible AI & FinOps | OWASP risks, regulation and cost control come up in every senior interview. | Week 9 |
| GenAI system design & leadership | Architects are judged on reference architectures, ROI and operating models. | Week 10 |
✅ Before You Start
Prerequisites, entry assessment and setup
Entry assessment (20 minutes): 15 questions on Python (6), SQL (4), APIs and JSON (3) and cloud basics (2). Score 60%+ to start directly; below that, you get a free 6-hour prep pack before Week 1. Professionals with 5+ years are placed in the Architect track by default. Python and SQL are not taught in the course.
What you should already be able to do
| Python | Write functions and classes, use pip / virtual environments, read JSON, call a REST API with requests |
|---|---|
| SQL | Joins, GROUP BY, window functions basics, CREATE TABLE |
| Git | Clone, commit, push, open a pull request on GitHub |
| Cloud | Log in to an AWS or Azure console, understand storage buckets and IAM roles at a basic level |
| Command line | Navigate folders, set environment variables, run a Python script |
🧰 Setup checklist (complete before Week 1)
- Databricks Free Edition (or trial) workspace, verified login
- Python 3.11+ with VS Code (or any IDE) and Git installed
- GitHub account with an empty repo for your course portfolio
- Anthropic and OpenAI API keys with a small prepaid credit (about US$10 each is enough)
- Node.js LTS (for MCP tools) and Ollama installed on your laptop
- Docker Desktop (optional, used in Weeks 8-9)
- n8n Cloud trial or self-hosted n8n via Docker (before Week 9)
📚 Detailed Weekly Syllabus
Generative AI & Agentic AI syllabus, week by week
Each week lists what you will learn, how it runs on Databricks, the core lab, the Architect-track extension, your deliverable and the interview questions it prepares you for. Click any week to expand.
WEEK01How Generative AI ThinksLLM Foundations, Reasoning Models & Open Models · 12 hours
By the end of this week, you can explain how LLMs work, pick the right model for a task, and call Claude, GPT and open models from code.
What you will learn
- How LLMs work: tokens, context windows, transformer intuition, next-token prediction
- Key settings: temperature, top-p, max tokens, stop sequences
- Reasoning models: extended thinking, effort levels, when reasoning is worth the cost
- Model landscape: Claude (Opus, Sonnet, Haiku), OpenAI GPT, Gemini, Llama, Mistral, Qwen, DeepSeek
- Calling models via Anthropic SDK, OpenAI SDK and OpenAI-compatible endpoints; token cost maths
- Open-weight and small language models (SLMs): Ollama, vLLM, quantization
- ◆ ArchitectEdge and on-device GenAI: quantized SLMs (GGUF, ONNX, llama.cpp) for laptops, mobile and IoT
- Responsible AI basics: hallucination, bias, privacy, licensing
🧱 Databricks lens
Foundation Model APIs (pay-per-token vs provisioned throughput), AI Playground, external model endpoints for Claude and GPT, workspace setup with Unity Catalog.
🧪 Core lab
Run the same task on Claude, GPT, a local Ollama model and a Databricks-served open model; compare quality, latency and cost.
◆ Architect track
Model selection framework: capability vs latency vs cost vs data residency; a multi-vendor model portfolio instead of betting on one provider.
DeliverableModel comparison notebook: quality, latency and cost for 4 models on the same task.
AssessmentCore: 10-question quiz + lab review · Architect: 1-page model selection policy for a company
Interview-ready topicsTokens vs words, why temperature 0 is not fully deterministic, reasoning vs non-reasoning models, open vs closed models.
WEEK02Talking to AI EffectivelyPrompt & Context Engineering · 12 hours
By the end of this week, you can write reliable prompts, get guaranteed structured JSON output, and run LLMs over millions of rows with SQL.
What you will learn
- System prompts, roles, zero / one / few-shot prompting
- Chain-of-thought, step-back and self-consistency prompting
- Structured outputs: JSON schema, Pydantic models, validation and retries
- Context engineering: selecting, compressing and ordering what goes into the window
- Prompt caching, batching and cost reduction
- Prompt injection (direct and indirect) and jailbreak defence
- ◆ ArchitectProgrammatic prompt optimisation with DSPy
- ◆ ArchitectGenAI inside data pipelines: idempotency, re-processing cost, prompt versions in lineage
🧱 Databricks lens
AI Functions in SQL: ai_query(), ai_classify(), ai_extract(), ai_summarize(); LLM steps inside Lakeflow pipelines; MLflow Prompt Registry.
🧪 Core lab
Classify and extract entities from 10,000 support tickets with ai_query() in a scheduled pipeline; then auto-optimise the prompt with DSPy.
◆ Architect track
Batch vs real-time inference design, and treating prompts as versioned, tested artefacts like code.
DeliverableTicket-classification pipeline writing structured results to a Delta table.
AssessmentCore: 10-question quiz + lab review · Architect: design note on batch LLM inference cost and reprocessing
Interview-ready topicsFew-shot vs fine-tuning, how to guarantee valid JSON, context vs prompt engineering, indirect prompt injection.
WEEK03Teaching AI Your Company's DataEmbeddings, Vector Search & RAG · 12 hours
By the end of this week, you can build a chatbot that answers from your own documents, including scanned PDFs, with citations.
What you will learn
- Embeddings: what they capture, embedding models, dimensions and cost
- Document AI: parsing PDFs, tables and scanned images with vision models
- Chunking strategies: fixed, recursive, semantic, contextual retrieval
- Vector stores: FAISS, Chroma, pgvector, Pinecone; ANN indexes (HNSW)
- Similarity search, metadata filtering, hybrid (keyword + vector) search
- The RAG pipeline end to end: ingest, chunk, embed, retrieve, generate, cite
- ◆ ArchitectIncremental indexing with CDC, document-level access control, re-indexing when models change
How a RAG pipeline works
Documents→Parse→Chunk→Embed→Vector Search→Retrieve→LLM→Answer + citations
🧱 Databricks lens
Databricks Vector Search (Delta Sync and Direct Access indexes), ai_parse_document() for files in Volumes, embedding endpoints, Unity Catalog permissions on chunks.
🧪 Core lab
Volumes → parsed Delta table → Vector Search index → answers with citations, over policy PDFs and scanned forms.
◆ Architect track
RAG at enterprise scale: millions of documents, freshness SLAs and row-level security in retrieval.
DeliverablePolicy-document RAG chatbot with citations on Databricks Vector Search.
AssessmentCore: 10-question quiz + lab review · Architect: RAG architecture for 5 million documents with access control
Interview-ready topicsChunk size trade-offs, when hybrid search beats vector search, how to enforce row-level security in RAG.
WEEK04Making AI Answers Accurate & TrustworthyAdvanced RAG, Knowledge Graphs & Evaluation · 12 hours
By the end of this week, you can measure answer quality with numbers, improve a RAG system step by step, and use knowledge graphs for connected questions.
What you will learn
- LangChain essentials: chat models, prompt templates, LCEL, retrievers, output parsers
- Advanced retrieval: query rewriting, multi-query, HyDE, parent-document, re-ranking
- Knowledge graphs + GenAI: extracting entities and relations with LLMs, GraphRAG, Neo4j
- Evaluation metrics: groundedness, relevance, correctness; LLM-as-a-judge; RAGAS
- Synthetic test data: generating realistic question-answer sets for evaluation
- Semantic caching for repeated questions
- ◆ ArchitectEvaluation-driven development: quality gates in CI, aligning LLM judges with human reviewers
🧱 Databricks lens
MLflow 3 for GenAI: tracing, evaluation datasets, built-in and custom LLM judges (scorers), Review App for expert feedback.
🧪 Core lab
Generate a 100-question synthetic eval set, baseline the Week 3 bot, apply re-ranking and query rewriting, add a small knowledge graph, and prove the gains.
◆ Architect track
When GraphRAG is worth its build cost, and how to stop quality regressions reaching production.
DeliverableEvaluation report in MLflow proving measurable improvement over the Week 3 bot.
AssessmentCore: 10-question quiz + lab review · Architect: release quality-gate policy for a GenAI product
Interview-ready topicsHow do you evaluate RAG in production? Explain re-ranking. Vector RAG vs GraphRAG. Weaknesses of LLM-as-a-judge.
WEEK05Building Your First AI AgentsAI Agents, LangGraph & Memory · 12 hours
By the end of this week, you can build an agent that uses tools, remembers users across sessions and asks for human approval before risky actions.
What you will learn
- What makes an agent: LLM + tools + memory + control loop (ReAct pattern)
- Workflows vs agents: routing, prompt chaining, orchestrator-workers, evaluator-optimizer
- Tool / function calling with Claude and GPT; designing good tools
- LangGraph: state, nodes, edges, conditional routing, checkpointers, human-in-the-loop
- Agent memory: short-term, long-term, episodic and semantic (LangMem, Mem0)
- ◆ ArchitectLong-running agents: context compaction, durable state, resume after failure
How an AI agent works
User goal→LLM plans→Calls a tool→Reads the result→Updates memory→Next step or answer
🧱 Databricks lens
Mosaic AI Agent Framework, Unity Catalog functions as governed tools, ResponsesAgent interface, Lakebase (managed Postgres) as agent state and memory store.
🧪 Core lab
A LangGraph agent that writes and runs SQL on Unity Catalog, remembers user preferences, and asks for approval before any write.
◆ Architect track
Choose the least-autonomous design that works; failure modes and cost blow-ups of open-ended agent loops.
DeliverableData Analyst agent over Unity Catalog tables with memory and approval steps.
AssessmentCore: 10-question quiz + lab review · Architect: agent vs workflow decision record for a real use case
Interview-ready topicsAgent vs workflow, how LangGraph manages state, memory design, how you stop an agent looping or misusing tools.
WEEK06AI Teams That Work TogetherMulti-Agent Systems: CrewAI, AutoGen, PydanticAI · 12 hours
By the end of this week, you can build multi-agent systems in three frameworks and choose the right framework for a job.
What you will learn
- Multi-agent patterns: supervisor, router, hierarchical teams, handoffs
- CrewAI: agents, roles, tasks, crews and flows
- AutoGen and Microsoft Agent Framework: conversational multi-agent systems
- PydanticAI: type-safe agents, structured results, dependency injection
- OpenAI Agents SDK and Claude Agent SDK: handoffs, sub-agents, built-in tools
- Framework selection guide: which one for which problem
- ◆ ArchitectWhen NOT to use multi-agent: token cost multiplication, coordination failures, debuggability
🧱 Databricks lens
Agent Bricks and multi-agent supervisor, Genie spaces as governed text-to-SQL agents, MLflow tracing for any framework.
🧪 Core lab
Build the same Market Research use case in CrewAI, PydanticAI and LangGraph; compare code size, reliability, cost and traces.
◆ Architect track
Framework standardisation for a team: one default, clear exceptions, and exit costs.
DeliverableMarket-research system built in 3 frameworks with a comparison scorecard.
AssessmentCore: 10-question quiz + lab review · Architect: framework standardisation proposal for a 20-person team
Interview-ready topicsCompare LangGraph, CrewAI and AutoGen. When would you choose PydanticAI? When is a single agent better?
WEEK07Connecting AI to Tools & Other AgentsAgent Protocols: MCP, A2A & Agent Skills · 12 hours
By the end of this week, you can build your own MCP server, plug it into Claude and your agents, and let agents from different teams work together.
What you will learn
- Model Context Protocol (MCP): host, client, server; tools, resources and prompts
- Transports (stdio, Streamable HTTP) and OAuth authentication
- Building MCP servers in Python (official MCP SDK / FastMCP)
- Agent2Agent (A2A) protocol: agent cards, tasks, cross-vendor collaboration
- Agent Skills: packaging reusable instructions, scripts and resources for agents
- Protocol security: tool poisoning, over-permissioned servers, secrets
- ◆ ArchitectEnterprise tool platform: internal MCP registry, on-behalf-of user identity, auditing every tool call
MCP connects agents to tools; A2A connects agents to agents
Your agent→MCP client→MCP server→Delta tables · Vector Search · Genie→A2A→Another team's agent
🧱 Databricks lens
Databricks managed MCP servers (Unity Catalog functions, Vector Search, Genie), hosting custom MCP servers on Databricks Apps.
🧪 Core lab
Build an MCP server for Delta tables and Vector Search; connect it to Claude Desktop and your Week 5 agent; expose the agent to another agent via A2A.
◆ Architect track
Governing hundreds of tools across teams without losing security or auditability.
DeliverableMCP server exposing Delta tables and Vector Search, used by Claude Desktop and your agent.
AssessmentCore: 10-question quiz + lab review · Architect: MCP governance model for an enterprise
Interview-ready topicsMCP vs function calling vs A2A, tools vs resources, how you secure a remote MCP server.
WEEK08AI That Sees, Hears, Browses & CodesMultimodal, Browser, Computer-Use & Coding Agents · 12 hours
By the end of this week, you can build agents that read images and documents, talk by voice, operate a browser, and speed up your own coding.
What you will learn
- Multimodal agents: image and document reasoning, charts and screenshots as input
- Voice agents: speech-to-text, realtime speech APIs, text-to-speech, latency budgets
- Browser automation: Playwright, browser-use library, Playwright MCP
- Computer-use agents: Claude computer use and OpenAI computer-use models; DOM vs vision
- Agentic coding tools: Claude Code, OpenAI Codex, GitHub Copilot, Cursor
- Ethics and site terms: robots.txt, credentials, CAPTCHA boundaries
- ◆ ArchitectSpec-driven development: project memory files, sub-agents, hooks and review loops for data code
🧱 Databricks lens
Scheduling agents with Lakeflow Jobs, landing data in Bronze Delta tables, secret scopes; coding agents with Asset Bundles and the Databricks CLI.
🧪 Core lab
Build a price-tracking browser agent that loads validated JSON into Delta; add an image-reasoning step; use a coding agent to write its tests and CI.
◆ Architect track
Sandboxing and approval gates for autonomous agents; measuring real productivity gains from coding agents.
DeliverableScheduled browser agent loading validated data into Delta, plus a voice or image-reasoning demo.
AssessmentCore: 10-question quiz + lab review · Architect: risk controls for a computer-use agent in production
Interview-ready topicsDOM vs vision automation, making browser agents reliable, voice-agent latency, trusting AI-written code.
WEEK09Running AI Safely in ProductionLLMOps, Security, Responsible AI & n8n · 12 hours
By the end of this week, you can deploy, monitor and secure a GenAI app, control its cost, and automate business workflows around it with n8n.
What you will learn
- LLMOps (3 hrs): packaging, CI/CD, serving, model routing and fallbacks, OpenTelemetry tracing, Langfuse / LangSmith
- Beyond Databricks (1 hr): Hugging Face, Kubernetes with vLLM, AWS Bedrock, Azure AI Foundry, Google Vertex AI
- AI security (2 hrs): OWASP Top 10 for LLM and agentic apps, red-teaming, guardrails (Llama Guard, NeMo Guardrails), PII masking
- Responsible AI & safety (1 hr): bias and fairness testing, alignment and interpretability basics, EU AI Act and India DPDP Act
- Fine-tuning (2 hrs): when to fine-tune, LoRA / QLoRA, synthetic training data pipelines, distillation
- n8n workflow automation (3 hrs): webhooks, schedules, AI Agent and API nodes, human approval, retries and error workflows
- ◆ ArchitectGenAI FinOps: cost per request, model cascading, caching strategy, chargeback per team
A production GenAI architecture on Databricks
Databricks Apps UI→AI Gateway→Model Serving (agent)→Vector Search · UC tools · Lakebase→MLflow traces & inference tables→Monitoring & cost dashboard
🧱 Databricks lens
Model Serving for agents, AI Gateway (usage tracking, inference tables, guardrails, rate limits), MLflow Model Registry in Unity Catalog, monitoring, Databricks Apps, Asset Bundles.
🧪 Core lab
Deploy your agent behind AI Gateway with a Databricks Apps UI and cost dashboard; red-team it with 20 attack prompts; trigger it from n8n with retries and approval.
◆ Architect track
SLOs for GenAI (latency, quality, cost), incident response for AI failures, and portability across clouds.
DeliverableDeployed agent with gateway, UI, monitoring, cost dashboard, red-team report and an n8n workflow.
AssessmentCore: 10-question quiz + lab review · Architect: SLO and incident runbook for a GenAI service
Interview-ready topicsHow do you monitor an LLM app? Cut inference cost by 50%? Top LLM security risks? RAG vs fine-tuning? n8n vs LangGraph?
WEEK10Designing AI Systems & Getting HiredGenAI Architecture, Capstone & Career · 12 hours
By the end of this week, you can design a GenAI system end to end, defend your trade-offs in an interview, and present a portfolio-ready capstone.
What you will learn
- GenAI system design: reference architectures for assistants, copilots and agentic workflows
- Build vs buy, ROI and TCO, choosing use cases that actually ship
- Capstone build and review: evaluation, security, reliability, cost
- Databricks Certified Generative AI Engineer Associate: exam blueprint and practice questions
- GenAI resume, GitHub portfolio, LinkedIn; mock system-design interviews
- ◆ ArchitectLeading GenAI adoption: team operating model, platform vs product teams, pilot to production
🧱 Databricks lens
Full Mosaic AI stack in one project: Unity Catalog, Vector Search, Agent Framework, Lakebase, MLflow evaluation, Model Serving, AI Gateway, Databricks Apps.
🧪 Core lab
Present your capstone live: architecture, demo, eval scores, red-team results and cost per 1,000 requests.
◆ Architect track
Executive-level communication: turning a technical design into a business case.
DeliverableCapstone project on GitHub, scored on the rubric, with a live demo.
AssessmentCore: 10-question quiz + lab review · Architect: GenAI roadmap and business case for a mid-size company
Interview-ready topicsDesign a GenAI assistant for a bank's customer support. Walk through your capstone trade-offs.
🏆 Capstone Project
Pick one capstone, build it from Week 6, present it in Week 10
It is judged on how well it is engineered, not only on whether it works.
📊
Text-to-SQL Data Analyst Agent
Natural-language questions over Unity Catalog tables with SQL validation, memory, charts and human approval.
🎧
Customer Support Multi-Agent System
Triage, knowledge lookup and reply-drafting agents under a supervisor, with MCP tools and n8n ticket integration.
🔎
Competitive Intelligence Agent
Browser agent that tracks competitor prices and news, lands data in Delta and sends a weekly AI summary.
🧾
Unstructured-to-Structured ETL
LLM pipeline that turns invoices, contracts or emails into governed Delta tables with quality checks.
🩺
Multimodal Claims Assistant
Reads claim forms, photos and voice notes, checks them against policy documents and drafts a decision for human review.
Scoring rubric (100 points)
| Criterion | Points | What we look for |
|---|---|---|
| Architecture | 20 | Clear design, right level of autonomy, justified tool and model choices |
| Evaluation quality | 20 | Eval set, metrics, before/after evidence in MLflow |
| Security & responsible AI | 15 | Red-team results, guardrails, access control, PII handling |
| Reliability | 15 | Error handling, retries, fallbacks, monitoring and traces |
| Cost | 10 | Cost per 1,000 requests measured, with at least one optimisation |
| Demo & communication | 20 | Working live demo, clear explanation of trade-offs |
Result: 60+ = Pass (course certificate) · 75+ = Merit · 85+ = Distinction and a trainer LinkedIn recommendation. Architect-track learners also submit an architecture decision record.
🧰 Tools & Technologies
GenAI tools and frameworks covered
Tools and model versions are updated each batch to match the latest releases.
| Models | Claude (Opus, Sonnet, Haiku), OpenAI GPT, Gemini, Llama, Mistral, Qwen, DeepSeek; Ollama, vLLM, llama.cpp |
|---|---|
| Frameworks | LangChain, LangGraph, CrewAI, AutoGen / Microsoft Agent Framework, PydanticAI, OpenAI Agents SDK, Claude Agent SDK, DSPy |
| Protocols | Model Context Protocol (MCP), Agent2Agent (A2A), Agent Skills |
| Retrieval, graphs & memory | Databricks Vector Search, FAISS, Chroma, pgvector, Pinecone, Neo4j, LangMem, Mem0 |
| Automation & multimodal | n8n, Playwright, browser-use, Claude computer use, Claude Code, Codex, Copilot, realtime speech APIs |
| Databricks | Unity Catalog, Foundation Model APIs, AI Functions, Lakeflow, Mosaic AI Agent Framework, Agent Bricks, Genie, Lakebase, MLflow 3, Model Serving, AI Gateway, Databricks Apps |
| LLMOps & security | MLflow, RAGAS, Langfuse / LangSmith, OpenTelemetry, Llama Guard, NeMo Guardrails, Asset Bundles, GitHub Actions |
| Other platforms | Hugging Face, Kubernetes, AWS Bedrock, Azure AI Foundry, Google Vertex AI |
💳 Fee & What You Get
₹50,000₹75,000Save ₹25,000
0% EMI available · 7-day money-back guarantee after the first live class
- Live instructor-led classes with recordings for revision
- Databricks labs, starter code and solution notebooks
- GitHub-ready portfolio: 9 lab projects + 1 capstone
- Weekly quizzes and Databricks GenAI Engineer Associate practice questions
- Architect design-note reviews for senior learners
- Resume review, LinkedIn tips and mock interviews
- Placement assistance and doubt-clearing support
Book your free GenAI demo class
Trainer Venu's team calls you back within 2 hours (9 AM – 9 PM IST).
👨🏫 Your Trainer
Learn directly from Trainer Venu
Venu Katragadda
Founder, Sreyobhilashi IT · 14+ years in Big Data & Cloud Data Engineering
Venu has trained 1200+ working professionals on Spark, Databricks, AWS and Azure, and still teaches every session himself — no junior trainers, no recorded-only classes. The GenAI program applies the same production mindset to AI: agents that loop, RAG answers that drift, prompt injection, and inference bills that explode, plus the interview questions that follow.
PySpark & Spark internalsDatabricks & Delta LakeAWS Glue · EMR · Kinesis
Azure ADF · FabricKafka & Structured StreamingAirflowCI/CD with DABsClaude, MCP & AI agentsRAG & Vector SearchDatabricks Mosaic AI
⭐ Google Reviews
What our Databricks students say about Trainer Venu
Real, verified reviews from data engineers who trained with Venu — on Databricks, AWS, Azure, PySpark and streaming.
4.9
★★★★★
Based on 200+ Google reviews
★★★★★
“Recently took Databricks classes with Venu to upskill in trending technologies, and the experience exceeded all expectations. While I initially sought guidance only on Databricks, Venu provided in-depth training across the entire ecosystem — AWS, Kafka, NiFi, Airflow and PySpark.”
AZ
Abhishek Zararia
Databricks · Cleared DE Professional Cert
✅ Verified Google Review
★★★★★
“I had a truly valuable experience with Venu's Spark training along with AWS & Azure Databricks Training. He is highly knowledgeable, and the sessions are very well structured with extensive hands-on coverage.”
DM
Dhevipriya Marimutbhu
AWS & Azure Databricks Training
✅ Verified Google Review
★★★★★
“This training has exceeded my expectations. Venu explains concepts clearly and uses hands-on examples that make the content easy to understand. I am learning a lot and would definitely recommend it.”
NL
Pataballa N V Lakshminarayana
Databricks Training
✅ Verified Google Review
★★★★★
“I recently completed the Data Engineering course on Databricks and AWS. Venu Sir delivers instruction at the next level, focusing on high-performance learning. He explains every concept clearly and thoroughly, with practical examples.”
MM
Mahaboob Mulla
Databricks & AWS Training
✅ Verified Google Review
★★★★★
“Venu sir has explained end to end streaming project, data cleansing, and parsing various source data. This has helped me in my project work. The explanation on Spark architecture and other key concepts helped me understand Spark deeply.”
NS
Nimisha Shah
Databricks Streaming
✅ Verified Google Review
★★★★★
“I had a truly valuable experience with Sreyobhilashi's AWS & Azure Databricks Training & Placement Program. Hands-on coverage of Spark, Kafka, Flink, NiFi, Airflow, Azure and Snowflake.”
NT
Naveen Kumar Tavva
AWS & Azure Databricks Training
✅ Verified Google Review
❓ FAQ
Generative AI course: frequently asked questions
Who is this Generative AI & Agentic AI course for?
Freshers and career changers who know Python and SQL, data engineers and analysts moving into GenAI, ML engineers adding agents and LLMOps, and tech leads or architects with 5-15+ years who design GenAI platforms.
Do I need machine learning or deep learning experience?
No. The course is applied GenAI engineering: prompts, RAG, agents, evaluation and deployment. You need working Python, SQL, Git and basic cloud skills. Python and SQL are not taught, but learners who score below 60% in the entry assessment get a free 6-hour prep pack before Week 1.
How long is the course and how are classes scheduled?
10 weeks and 120 hours in total: 80 hours of live instruction plus 40 hours of guided labs and capstone work. Each week has 4 weekday live sessions and 1 weekend lab. Morning, evening and weekend batches are available, and every class is recorded.
What is the difference between the Core track and the Architect track?
Both tracks sit in the same batch. The Core track covers everything needed to build and ship GenAI apps. Professionals with 5+ years are placed in the Architect track by default, which adds advanced topics each week (reasoning models, enterprise RAG, FinOps, GenAI system design) and a 1-page design note reviewed by the trainer.
Does the course cover Databricks Mosaic AI?
Yes, every week has a Databricks lens: Foundation Model APIs, AI Functions such as ai_query(), Vector Search, Mosaic AI Agent Framework, Agent Bricks, Genie, Lakebase, MLflow 3 evaluation and tracing, Model Serving, AI Gateway and Databricks Apps.
Which certification does it prepare me for?
The Databricks Certified Generative AI Engineer Associate exam. Week 10 covers the exam blueprint and practice questions, and weekly quizzes include certification-style questions.
What will I have in my portfolio at the end?
9 weekly lab projects and 1 production-style capstone on your GitHub, including a RAG chatbot over your own documents, a SQL data-analyst agent with memory, your own MCP server used by Claude, browser, voice and multimodal agents, and a secured, monitored production deployment.
What is the course fee and refund policy?
The course fee is INR 50,000, with 0% EMI available. There is a 7-day money-back guarantee after the first live class, and you can attend a free demo class before you pay.
Do you help with jobs after the course?
Yes. Week 10 includes GenAI resume and LinkedIn reviews, GitHub portfolio preparation and mock system-design interviews, followed by placement assistance and doubt-clearing support.
📖 Glossary
Generative AI and Agentic AI terms in plain English
The key terms you will use throughout the course, explained simply.
- A2A (Agent2Agent)
- Open protocol that lets agents built by different teams or vendors discover each other and share tasks.
- Agent
- An LLM that works in a loop: it plans, calls tools, looks at the results and decides the next step.
- Agent Skills
- Reusable folders of instructions, scripts and resources that an agent loads only when a task needs them.
- AI Gateway
- A layer in front of models that controls access, rate limits, guardrails, routing and usage logging.
- Chunking
- Splitting documents into small pieces so the right piece can be found and passed to the LLM.
- Context window
- The maximum amount of text (in tokens) a model can read at once.
- Distillation
- Training a small model to copy a large model's behaviour, making it cheaper and faster.
- DSPy
- Framework that improves prompts automatically using examples and a metric, instead of manual trial and error.
- Embedding
- A list of numbers that represents the meaning of text, so similar meanings are close together.
- Fine-tuning
- Further training a model on your own examples to change its style or behaviour.
- GraphRAG
- RAG that uses a knowledge graph of entities and relationships to answer connected questions.
- Guardrails
- Checks on inputs and outputs that block unsafe, off-topic or sensitive content.
- Hallucination
- When a model states something confidently that is false or unsupported.
- Hybrid search
- Combining keyword search and vector search for better retrieval.
- Lakebase
- Databricks' managed Postgres database, useful for agent state, memory and app data.
- LLM
- Large Language Model, e.g. Claude or GPT: a model trained to predict and generate text.
- LLM-as-a-judge
- Using one LLM to grade another LLM's answers against criteria.
- LoRA / QLoRA
- Cheap fine-tuning methods that train small adapter layers instead of the whole model.
- MCP (Model Context Protocol)
- Open standard for connecting AI apps to tools and data through servers.
- MLflow
- Open-source platform for tracking, evaluating, tracing and deploying ML and GenAI apps.
- Model routing
- Sending each request to the most suitable model, e.g. a cheap model for easy questions.
- n8n
- Visual workflow-automation tool with AI nodes, used to connect agents to business apps.
- Prompt injection
- An attack where text in a prompt, document or web page tries to override the model's instructions.
- Quantization
- Storing model weights with fewer bits so the model runs on smaller hardware.
- RAG
- Retrieval-Augmented Generation: fetching relevant documents and giving them to the LLM to answer from.
- RAGAS
- Open-source library of metrics for evaluating RAG systems.
- Re-ranking
- A second, more accurate step that re-orders retrieved chunks by relevance.
- Red-teaming
- Deliberately attacking your own AI system to find weaknesses before others do.
- Semantic caching
- Reusing a previous answer when a new question means the same thing.
- SLM
- Small Language Model: compact model that can run locally or on devices.
- SLO
- Service Level Objective: a target such as '95% of answers in under 3 seconds'.
- Token
- The unit an LLM reads and writes; roughly three-quarters of an English word.
- Unity Catalog
- Databricks' governance layer for data, models, functions and AI tools.
- Vector Search
- Finding items by meaning using embeddings; Databricks offers a managed version.
Start building real AI agents on Databricks
Live online, 10 weeks, 120 hours. Morning, evening and weekend batches. Attend a free demo class before you pay.
Also explore: Databricks Data Engineering · Claude & Agentic AI Masterclass · AWS Data Engineering Syllabus