If you're a student in India and you're trying to pick which AI skills to focus on, the wrong answer is 'all of them.' The right answer is a layered approach: nail Tier 1 first, only then move to Tier 2, and treat Tier 3 as optional unless you're going into research.
Tier 1 — must learn (1 month)
- Prompt engineering — structured prompting, role assignment, chain-of-thought, output formatting
- AI tool fluency — Claude, ChatGPT, Gemini, Perplexity, daily use as a knowledge worker
- Vibe coding — building deployable apps using AI as your pair-programmer (Cursor, Bolt.new, Lovable)
- No-code automation basics — n8n or Zapier, build your first 3 workflows
Tier 2 — high value (2-3 months)
- RAG (Retrieval-Augmented Generation) — making AI work with private documents
- Fine-tuning intro — adjusting model output to match a brand voice or specific tone
- AI agents — multi-step automations that decide their own next action
- API integration — wiring AI into apps via Claude API, OpenAI API, Cloudflare Workers AI
- Vector databases — Pinecone, Supabase pgvector, Weaviate basics
Tier 3 — only if going into research
- Linear algebra and calculus for ML
- Neural network architecture (transformers, attention, embeddings)
- Model training from scratch in PyTorch / JAX
- Reading and reproducing AI papers
Why most students get this wrong
Most students start with Tier 3 because school curricula are stuck there. They drop out before reaching Tier 1. The result: a year of math homework and zero shipped projects. The order matters.
Where ONROL fits
ONROL's Generalist track covers all of Tier 1 and the practical half of Tier 2 in 5 days, with cohort-based mentorship. The Orchestrator track covers the rest of Tier 2 in depth. Tier 3 is best learned at IITs, IISc, or via specialised programs — ONROL doesn't try to cover that ground.