Build a RAG Knowledge Assistant
Grounded Answers You Can Evaluate
By Day 5: Build one RAG assistant with source citations and an evaluation showing retrieval and answer quality on a held-out question set.
Build a RAG Knowledge Assistant is a five-day live online WorkReady Capability Sprint for technical freshers, analysts moving into AI engineering, developers and GCC aspirants who want hands-on generative-AI application experience. Days 1–4 are 90-minute sessions and Day 5 is a 2–3 hour Capability Showcase; the complete Sprint costs ₹2,000.
- ₹2,000
- Complete 5-day Sprint
- Days 1–4
- 90 min/day live
- Day 5
- 2–3 hr Capability Showcase
Paid professional capability development. Not a job or placement service.
At a glance
- Who it's for
- Technical freshers, analysts moving into AI engineering, developers and GCC aspirants who want hands-on generative-AI application experience.
- Entry level
- Required: basic Python (functions, lists/dicts, pip, running a notebook). Readiness check before enrolment. Entry gate: Python — a short readiness check before enrolment. Learners who are not ready are routed to Python for Data Work, not admitted with hidden homework.
- Format
- Online · a focused five-day live online studio
- Duration and schedule
- Days 1–4 are 90-minute live sessions. Day 5 is an extended 2–3 hour Capability Showcase.
- What you produce
- RAG app + retrieval/answer evaluation
- Fees
- ₹2,000 for the complete 5-day Sprint
- What it is not
- Not a five-day path to "become an AI engineer". One evaluated RAG application. Not a job or placement service. SkillsNCR does not guarantee employment, placement, salary or interview outcomes.
Learners wire up a vector database from a tutorial but cannot explain chunking choices, check whether retrieval found the right passage, or measure hallucination.
Build a small retrieval-augmented (RAG) assistant over a defined document set, and prove — with an evaluation — when its answers can and cannot be trusted.
Your five days, day by day
Days 1–4 are 90-minute live sessions. Day 5 is an extended 2–3 hour Capability Showcase. Open each day to see what you do and what you leave with.
Day 1Problem, corpus and baseline90 min
- Use case and users (e.g. HR policy assistant, product manual Q&A, SOP lookup)
- Curate the corpus; remove sensitive data
- Write 20 test questions with expected source passages
- Baseline: LLM without retrieval — record failures
You leave the day with: Use-case brief + test set + baseline results
Day 2Ingest, chunk and embed90 min
- Parsing PDFs/HTML/Docs; cleaning
- Chunk size/overlap trade-offs; metadata
- Embeddings and similarity search
You leave the day with: Indexed corpus + retrieval-only test results
Day 3Generate grounded answers90 min
- Prompt that forces grounding and citation
- "I don't know" behaviour when retrieval is weak
- Minimal UI
You leave the day with: Working RAG v1
Day 4Evaluate and improve90 min
- Retrieval hit-rate on test set; answer faithfulness and correctness scoring
- Failure analysis: wrong chunk, missing doc, conflicting sources, prompt injection inside documents
- One improvement cycle (chunking, metadata filter, re-ranking or query rewrite)
- Human review and revision
You leave the day with: Evaluation report before/after + RAG v2
Day 5Capability Showcase2–3 hr
Changed-case challenge: Reviewer adds a new/updated document and new questions
- Demo with citations
- Learner re-ingests, re-evaluates and explains the change in scores
- Discuss cost, privacy and when not to use RAG
You leave the day with: Final proof pack, submitted after the changed-case challenge
Tools you work with
- Python
- Jupyter / Colab
- An LLM API
- Embeddings model
- Chroma / FAISS or pgvector
- LangChain / LlamaIndex (optional)
- Streamlit / Gradio
Tool names describe what is used in the Sprint. No vendor endorsement, certification or licence is implied.
RAG app + retrieval/answer evaluation
This is the work you explain, question and adapt in the Day 5 Capability Showcase.
Your proof pack
- 01Use-case brief
- 02Corpus inventory
- 03Test set
- 04Code/notebook
- 05Retrieval and answer evaluation (before/after)
- 06Failure analysis
- 07Demo recording
- 08Limitations note
What a reviewer checks
- Answers cite retrievable sources
- Assistant declines when evidence is missing
- Evaluation is reproducible
- Updated corpus handled correctly
- Learner explains chunking/retrieval trade-offs
Your work has to survive a changed case.
On Day 5 you present the work, explain key decisions, identify limitations and respond when part of the case changes. The final proof pack is submitted after that challenge.
- Present
Walk the reviewer through the work and the decisions behind it.
- Question
Answer challenges on evidence, assumptions and risk.
- Adapt
Reviewer adds a new/updated document and new questions
- Defend
Submit the final proof pack and explain what changed and why.
₹2,000for the complete 5-day Capability Sprint
- Days 1–4
- Live Sprint sessions · 90 minutes/day
- Day 5
- Extended Capability Showcase · 2–3 hours
Paid professional capability development. Not a job or placement service.
Included in the Sprint
- An applied professional case
- Human review and feedback
- Revision after feedback
- A Day 5 changed-case challenge
- A final proof pack: RAG app + retrieval/answer evaluation
Questions worth asking.
Is a Capability Sprint just a short course?
No. It is a focused live capability studio built around doing one bounded kind of professional work, receiving feedback, revising it and demonstrating what you can do.
Will this make me an AI engineer?
No. The Sprint builds one bounded capability: build one RAG assistant with source citations and an evaluation showing retrieval and answer quality on a held-out question set. Not a five-day path to "become an AI engineer". One evaluated RAG application. SkillsNCR does not promise a role, job or placement.
Do I need prior experience?
Required: basic Python (functions, lists/dicts, pip, running a notebook). Readiness check before enrolment. Entry gate: Python — a short readiness check before enrolment. Learners who are not ready are routed to Python for Data Work, not admitted with hidden homework.
What will I produce?
Use-case brief, Corpus inventory, Test set, Code/notebook, Retrieval and answer evaluation (before/after), Failure analysis, Demo recording, Limitations note.
What does a Sprint cost?
₹2,000 for the complete five-day Sprint.
How long is it?
Days 1–4 are 90-minute live sessions. Day 5 is an extended 2–3 hour Capability Showcase.
When is the next cohort?
This Sprint runs on request — for a group, an employer or a college, or when waitlist demand justifies it. Request it on WhatsApp for yourself or your team.
Can I join only one Sprint?
Yes. Each Sprint is designed to provide standalone value.
Does this guarantee a job or placement?
No. WorkReady is paid professional capability development, not a job or placement service. SkillsNCR does not guarantee employment, placement, salary or interview outcomes.
Request Build a RAG Knowledge Assistant for yourself or your team.
Paid professional capability development. Not a job or placement service.