Blogs/Posts
From Models to Agents: The Missing Layer Between AI and Real Problems
Breaks down the gap between a raw LLM and a system that can actually act on the world. Covers how RAG grounds model outputs in real data, how agents chain reasoning and tool use, and where MCP fits as the protocol layer connecting it all. Built around a running example so the abstractions stay concrete. Featured by Towards AI.
Emerging Threats in LLMs
A Fun Friday session I led at Zoho covering how LLMs fail under adversarial pressure. Real incidents in each: prompt injection, data poisoning, model theft, privacy leaks and how they mapped to the OWASP Top 10 framework, with red-teaming practices from OpenAI, DeepMind, and Anthropic.
The Problem Chain That Led to Transformers
A deep dive into the sequence of problems and limitations in earlier architectures — RNNs, LSTMs, attention — that collectively motivated the design of the Transformer. Traces the intellectual journey that shaped modern NLP.