Skip to main content

Experience

Slice Bank·Sep 2024 — Present

Engineering Manager

3 squads · 15+ engineers·Builder + Manager

Slice became a bank in late 2024 with no merchant product. Built 4 live payment API product lines across merchant banking, payment APIs, and digital lending.

Key outcomes

  • Built Slice's PA/PG merchant business from scratch — 1 customer and <₹5L/month to 4 live product lines (Account Validation, Payouts, PayIns, Mandates) generating ₹70L+/month in 15 months
  • PayIn P99 latency: 200s → 7.5s (15×) — rebuilt architecture under live production traffic while simultaneously scaling the product
  • AI Oncall Bot: 71% query hit rate, 90s avg resolution vs. 5–10 min baseline — adopted org-wide beyond my squads
  • AI OpEx Reporter: 4h/week per squad saved via automated CloudWatch/Sentry → Confluence/Slack pipeline
  • Managerial AI Toolkit: 6h/week off EM calendar, 70% retro prep saved across the org
  • Cortex — multi-tenant RAG platform shipped at SliceLab hackathon; productionisation underway

The call

The decision to rebuild PayIn was a product and business call — not mine. What I owned was the engineering: running a full architecture overhaul under live production traffic, no downtime, no rollback, while the team was simultaneously building new product lines. That kind of execution — high-stakes, constrained, no clean room — is where I operate best. P99 went from 200s to 7.5s. No major incidents during the migration.

Tech & skills

Platform EngineeringPaymentsAI SystemsReliabilityFintech

JumpingMinds AI·Jun 2021 — Aug 2024

Co-Founder & CTO

2 founders → 15 engineers·Founder + Builder

Built an AI mental health platform from scratch before LLMs were mainstream. Grew to 1M+ users, then wound it down without wasting investor money chasing a broken model.

Key outcomes

  • Snuggles AI companion shipped early 2022 — months before ChatGPT (Nov 2022); datable, production AI system in a regulated consumer context
  • 1M+ users across India, US, UK, Scandinavia — 35 min daily engagement vs. 10–15 min industry average
  • Chat API P99: 3s → 250ms (12×) rebuilt under live traffic as user base scaled
  • $1.3M raised · Google for Startups Accelerator 2022 · backed by ex-Disney India CEO, CTO, and SVP Product
  • Pivoted to B2B corporate wellness in 2023 funding winter — extended runway by 14 months without external capital

The call

During the 2023 funding winter we were offered a 51% JV acquisition — the best deal available for the next 12 months. We said no: couldn't respect the acquirer's track record, and freedom to operate was non-negotiable. Chose a clean wind-down in August 2024 over the wrong partnership. The B2C unit economics had peaked; the right move was acknowledging it, not chasing the deal.

Tech & skills

AI / LLMPython / DjangoConsumer Product0 → 1HIPAA

Disney+ Hotstar·Jan 2018 — Jun 2021

Engineering Lead

15 engineers · Social, Gaming & Sports·Tech Lead + EM

Live cricket at 50M+ concurrent users — sub-second latency — and no off-the-shelf product that could handle the fan-out. Built the real-time messaging infrastructure from scratch.

Key outcomes

  • PubSub/MQTT infrastructure scaled from 35M → 50M+ CCU at <1s latency — 250B+ messages during the 2019 cricket season
  • Architecture held across 5 cricket seasons without rearchitecting — built right the first time
  • Co-founded the Architecture Council with Ashutosh Agrawal (now Staff SSE, Google DeepMind) — cross-team design reviews across 15+ teams, backed by CTO's office
  • SEA localisation platform — 80% reduction in operational effort per new country launch
  • RootConf Delhi 2020: public technical talk on the real-time infrastructure (40-min deep dive)

The call

Every commercial PubSub product we evaluated hit a ceiling at live sports concurrency — the fan-out characteristics of 50M users reacting to the same ball in the same second had no prior art. Built the infrastructure from scratch rather than adapting something that would fail at scale. That same architecture ran the 2020, 2021, 2022, and 2023 cricket seasons — no structural changes required.

Tech & skills

Real-time SystemsPubSub / MQTTConsumer ScaleArchitectureSEA Expansion

HyperTrack·Jun 2016 — Dec 2017

Lead Software Engineer

Small eng team · 1000+ developer community·Builder

Location intelligence platform for real-time movement tracking. No developer SDK ecosystem existed for mobile location — built and shipped the open-source Android libraries from scratch, owned engineering delivery end-to-end, and grew a developer community that put HyperTrack on the map.

Key outcomes

  • smart-scheduler-android: 727 GitHub stars, 20K+ developers — #1 GitHub Trending (Java) for 3 consecutive days, 256 stars in a single day
  • live-app (Android + iOS): 103 HN points, 40 comments, 2000+ developers — Show HN, zero marketing spend
  • hyperlog-android: 698 GitHub stars — mentored Aman Jain (junior mobile engineer) who co-built it; adopted as a utility logging library across the developer community
  • Android Weekly #234 feature — international developer community reach
  • Galileo Hackathon, Gdańsk, Poland — led HyperTrack's participation as the European developer signal
  • Cultivated a 1000+ developer community through open-source, technical blog, and ecosystem events; coached 2 senior engineers

The call

The product was B2B — enterprise location intelligence. The bet was open-sourcing the Android building blocks anyway, instead of keeping them closed. smart-scheduler-android hit #1 on GitHub Trending (Java) without a PR campaign — stayed on the trending page for 3 consecutive days, pulling 256 stars in a single day. That's the kind of credibility a startup can't buy: 20K+ developers showing up because the library solved a real problem, not because of outreach.

Tech & skills

AndroidMobile SDKsOpen SourceLocation TechJavaDeveloper Tools

Earlier career

2015 — 2016

Shuttl

Senior Software Engineer

Bus aggregation platform — early backend and routing APIs.

Want to talk about this work or explore how I can help your team?

Book a call