Hi, I'm Ashish Goyal

I'm a Machine Learning Engineer specializing in AI Agents and Retrieval-Augmented Generation (RAG). Currently at Verloop.io, I'm shaping the team's agent-platform strategy and building production RAG systems that power autonomous customer support for enterprise clients across e-commerce, banking, and healthcare.

I work hands-on with modern agentic frameworks — Claude Agent SDK, LangGraph, and Deep Agents — alongside Model Context Protocol (MCP), function calling, structured outputs, sub-agent orchestration, human-in-the-loop approval flows, and agent observability. On the retrieval side, I design evaluation frameworks for embedding models, cross-encoder rerankers, and late-interaction models, with production pipelines for complex document parsing, Weaviate vector search, and context engineering.

I also believe in understanding AI systems at a fundamental level. Through my technical blog, I break down topics like attention mechanisms (MHA, GQA, MLA), KV-cache, the LLaMA-2 architecture, and Mixtral 8×7B MoE, providing complete PyTorch implementations from scratch. I focus on building with clarity, sharing knowledge openly, and making advanced ML techniques accessible to practitioners.