I work on production LLM agents. At Laiqa I own Anya, a WhatsApp health companion (~1000 users) with evals and outbound rails. Before that I shipped calling infrastructure at Zenbase for 10k+ concurrent conversations. Mech eng at NIT Jalandhar, 2027.
Currently I'm working on context virtualization: a way for agents to have infinite context. That's contextd.
Layr Turns Slack, Jira, and Linear into a product knowledge graph, then drafts specs and tasks from it.
Design Agent Multi-agent CAD pipeline: natural language in, headless FreeCAD geometry out.
Hermes Agent Lazy MCP startup from an on-disk schema cache. Idle dashboard RSS from ~1GB to ~110MB.
nn-Scratch
C++ tensor/autograd core with a Python API. Built it to see what .backward() actually does.
Also Apache Superset and Cal.com.