Ayush Ranjan
Experience Projects Research
Experience Projects Research
Back to Research

AI

Models, agents, and tooling.

  • Agentic loops and tool use
    The think→act→observe cycle, how tool calling works, and when a loop beats a single prompt.
  • Retrieval-augmented generation in practice
    Chunking, embeddings, reranking, and the failure modes that quietly wreck RAG quality.
  • How transformer self-attention actually works
    Q/K/V projections, causal masking, and why the KV-cache makes generation fast.
© 2026 • Ayush Ranjan
Press Esc or click anywhere to close