Himshikhar

overview.md

Himshikhar is a two-month residential program at IIT Mandi. My track is Agentic AI: a mix of professor-led lectures, hands-on labs, industry-mentor sessions, daily quizzes, and projects — a full immersion in how modern agentic systems are actually built and reasoned about.

curriculum.md

Ground covered so far spans the fundamentals up to current practice.

• Deep learning fundamentals
• Architectures — FCNN, VAE, RNN, LSTM, Transformers
• Agentic AI fundamentals & framework basics
• LLM reasoning
• Retrieval-augmented generation (RAG)

mini-project.txt

Current mini-project: structured LLM workflows — designing reliable, composable pipelines around language models rather than treating them as a single opaque call. The recurring theme of the program: making LLMs dependable enough to act, not just answer.

lessons-learned.md

• Agentic reliability is an engineering problem, not just a prompting one.
• Structure around an LLM (tools, steps, checks) buys more than a bigger model.
• Daily quizzes are an underrated way to make fundamentals stick.

future-work.md

• Take structured-workflow patterns back into ExplainMyRepo and EmotionFlow.
• Go deeper on evaluation for agentic systems.
• Write up the program as a retrospective essay.