a production-grade GenAI application with retrieval
Your training passport
Every domain runs the same three-stage structure. Here's what each stage looks like for Generative AI.
Prompt engineering and evaluation
Weeks 1-2. You write 30+ prompts, build an eval harness, and learn to tell the difference between a model that's flaky and a model that's actually bad. Mentor scores your evals against their own.
1–2
Retrieval and grounding
Weeks 3-4. RAG over a real corpus, chunking strategy, embedding choices, source citations in the UI. You build a chatbot for a partner's internal docs and they actually try to break it.
3–4
Fine-tune or ship to production
Months 2-3. Either fine-tune a small open-weight model on a focused task, or ship a streaming chat product with cost-aware architecture (caching, batching, model selection). Both paths require a writeup.
2–3
- Prompt engineering & few-shot design
- RAG pipelines
- Fine-tuning open-weight models
- Evaluation of LLM outputs
- Streaming chat UX
- Cost-aware architecture
Once you're enrolled as an intern in Generative AI, two domain-matched courses unlock on your dashboard, alongside the two courses free to every intern.
Browse the course catalogReady to start?
Applications are open all year — there's no fixed intake date. Onboarding starts within two weeks of approval.
Apply for Generative AI