AI Engineering with Generative AI & Cloud
A project track covering a Python ML refresher, LLM fundamentals and prompt engineering, retrieval-augmented generation over your own documents, tool-using agents with evaluation, and guarded cloud deployment.
A real sequence, not a playlist
Six numbered weeks. Every week ends in a deliverable a reviewer actually checks — plan roughly twenty focused hours per week.
Course workspace: 23 lessons · 11 reads · 6 checkpoints · 6 reviewed builds
- 01~20h effort
Week 1 · Python & ML Refresher for LLM Work
- —Python environments, NumPy and pandas refresh
- —Train-test discipline and metric reporting refresher
- —API-key hygiene and cost-aware experimentation
Deliverable: Notebook reproducing a baseline classifier with metric table
- 02~20h effort
Week 2 · How LLMs Work & Prompt Engineering
- —Tokens, context windows, and temperature intuition
- —Instruction patterns, few-shot, and structured outputs
- —Failure modes: hallucination, drift, and prompt injection basics
Deliverable: Prompt pack with before-and-after outputs on a fixed task
- 03~20h effort
Week 3 · Embeddings & Retrieval-Augmented Generation
- —Embeddings, chunking, and vector-search literacy
- —Grounding answers in your own document set
- —Citations and "I don’t know" discipline
Deliverable: RAG Q&A over your own documents with cited answers
- 04~20h effort
Week 4 · Agents, Tool Use & Evaluation
- —Function calling and multi-step task decomposition
- —Eval sets and scoring model behaviour honestly
- —Hallucination checks and human-in-the-loop points
Deliverable: Tool-using agent demo with an eval scorecard
- 05~20h effort
Week 5 · Cloud Deployment & Guardrails
- —Wrapping the pipeline as an API with a Streamlit UI
- —Rate limits, content filters, and usage-cost tracking
- —Logging prompts and outputs for auditability
Deliverable: Deployed demo URL with guardrail notes and cost log
- 06~20h effort
Week 6 · Final Project Review
- —Project reflection on retrieval and agent decisions
- —Completion review against the published rubric
- —Training-completion record when requirements are met
Deliverable: Final project submission
What you actually build
No copy-paste tutorials or toy repos. Every enrolled student submits a production-grade portfolio project evaluated against an industry rubric.
Grounded Campus Assistant with RAG & Guardrails
LLM assistant that answers from your own documents: chunked corpus, vector retrieval with citations, a tool-using agent demo with an eval scorecard, and a deployed guarded demo.
Portfolio Deliverables Checked by Reviewers
Checked by named reviewers
Reviewers are named at cohort start and published with the batch. They check every submission against the rubric — passing names appear on the certificate.
- Every submission checked against a published academic and technical rubric
- One fair remediation attempt if a deliverable falls short of acceptance threshold
- No certificate without passing code checks — payment never substitutes genuine review
- Cryptographic verification QR permanently linked to your individual submission records
Where CSE / AI-DS / BCA / B.Sc students take this track
Matched by branch against the BEU college directory — not an affiliation claim. Academic credit is decided by each college's T&P cell, so ask your HoD to review the syllabus before you enrol.