Hi, I'm Yash. Press ⌘K to wander.

I build the systems behind products that have to work.

Senior full stack developer on the R&D team at Twilio Segment, shipping AI features to customers. 8 years of crawlers, data pipelines and Kubernetes platforms, and now agents, evals, and the generative UI protocols that put them in front of real users.

↑ go on, poke it. Click to scatter.

Twilio · Segment R&D
Twilio's first customer-facing AI product

Functions Copilot: plain-English prompts to working Segment functions. Hackathon demo to production: retrieval, a LangGraph agent, a streaming editor integration, and the eval set that made every change measurable.

0×
faster crawling after dropping Selenium
0%
lower infra bill at Zivame on spot Kubernetes
3.5h → 0m
Zivame nightly SKU sync after a concurrency rewrite
0+
years shipping across four companies

Built in the open

Two public projects on GitHub that people have starred.

2024★ 17

ManasAI

An open-source AI software engineer. It plans, writes and runs code in a Docker sandbox while you watch.

TypeScriptGraphQLDockerOpenAIAgents
2018★ 20

Amazon crawler

A Scrapy crawler for Amazon categories, product listings and product detail pages, including ratings and reviews.

PythonScrapyCrawling

The stack, in motion

Events and documents go in. An agent with tools and an eval gate does the thinking. AG-UI streams it, A2UI draws it, and a person stays in the loop.

Explore the stack →
production · bangalorelive
Eventsweb · mobile · APIDocumentsknowledge basePeopleapprove · editKafkastreams · topicsVector storeembeddings · RAGToolsMCP serversEvalsjudge · CI gateAgentlanggraph · modelsAG-UIevent streamGenerative UIA2UI · react

Have a system that has to work?

I'm happiest on problems with real constraints: throughput, cost, correctness, or an agent that needs to behave. Say hello.

Get in touch