Building at the edge of AI & product

I build things that think, and think about things worth building.

I work where AI meets real products, turning fuzzy problems into tools people actually use. Lately that means semantic search, agentic systems, and open source. Always learning in public.

Scroll
01

About

I'm a builder who's equally at home in a codebase and a product spec.

My favourite problems live in the gap between what's technically possible and what people actually need. I like shipping small, learning fast, and writing things down so others can build on them.

Most of my recent work orbits applied AI: semantic code search, retrieval systems, and agentic tooling, plus open source, where I contribute to projects like CocoIndex. When I'm not coding I'm usually thinking about how to make a product feel inevitable.

Off-screen: I won Smart India Hackathon 2024 (1st of 600+ teams), placed Top 5 at IIM Mumbai's ESG Hackathon, and led an 80+ member community NGO. I care about momentum, taste, and doing the unglamorous work that makes things real.

NowBuilding & shipping
FocusAI · Product · OSS
BasedIndia · open to remote
StudiedCS · Chandigarh University
StackPython · Postgres · pgvector · AWS
02

How I work

Think in systems
I map the whole flow before touching the parts: where things break, where users drop, where leverage hides.
Build with AI
Retrieval, embeddings, agents. I use AI as a material, not a buzzword, and ship tools that hold up locally.
Learn in the open
Open source, writing, and small public experiments. The fastest way to get good is to share the messy middle.
03

Work

Signzy
Dec 2025 · Present · Fintech
Associate Business Analyst

I work on Signzy's customer onboarding product, sitting between product, compliance, engineering, and integration teams to ship changes that are both usable and regulatory-compliant.

  • Revamped the RBI-compliant onboarding journey to reduce its length by 50%, and authored the PRDs that defined the new flow.
  • Built signature-matching, document-identification, and Board Resolution OCR APIs powered by LLM models, replacing slow manual verification steps.
  • These onboarding APIs are used by HDFC Bank to onboard millions of customers, reducing FTNR and easing the overall process.
  • Tracked onboarding funnel metrics and reduced drop-offs by 30% by identifying the key bottlenecks driving fall-off.
  • Owned sprint tracking, issue resolution, and cross-functional coordination on Jira across product, compliance, engineering, and integration teams.
OnboardingLLM OCR APIsPRDsJiraRBI complianceHDFC Bank
Medoc Health
Mar 2025 · Aug 2025 · Healthtech
Product Management Intern · Founder's Office

Doctors were hesitant to use the app for consultations, so I worked on the doctor-side application to find and fix the friction.

  • Introduced scribble and voice features that made writing prescriptions effortless for doctors during a consult.
  • This increased doctor usage of the app and lifted the overall consultation rate on the platform (weekly active doctors up ~25%).
  • Ran the discovery, defined success metrics, and shipped iteratively with engineering while keeping it clinically usable.
DiscoveryVoice & scribble0→1Adoption
04

Projects

05

Writing

LLMs alone are not the future You can't discover something genuinely new by reshuffling what already exists. Why the next leap is LLMs and reinforcement learning together. You have to be unemployed to keep up with the AI space right now New models every week, new agents, new stacks. The only thing that makes sense is staying curious and running small experiments.

Let's talk

Got a problem worth
solving? Say hello ↗