About

Shreyas Ponkshe

Lead engineer at Velotio (R Systems), currently on a Netflix engagement. Backend and cloud for a decade, now spent on production GenAI. Twice a co-founder.

Personal

I'm a backend engineer and a two-time co-founder, based in Pune, India. I mostly write Node.js, TypeScript, and Python, and I have spent most of the last ten years somewhere between an API and the infrastructure underneath it.

For the last few years that has meant production GenAI. The GenAI part is less a pivot than the same job with a new failure mode: retrieval and agents are still systems that have to hold up under load, cost money to run, and be observable when they go wrong.

Co-founding two companies means I have owned the infrastructure bill alongside the architecture. Running a compute-heavy pipeline on a startup budget set most of my instincts about what to build and what to buy.

What I Spend Time On

  • Backend systems and APIs in Node.js and TypeScript
  • Production GenAI: agents, RAG, and LLM assistants people use daily
  • Retrieval pipelines, prompt budgets, and evaluation against real traffic
  • Event-driven architecture and data pipelines
  • Cloud architecture across AWS, GCP, and Azure
  • Engineering leadership: growing a team, delivery, and review standards
  • The founder side of it: infrastructure cost, tradeoffs, and a small team

Professional

Lead Engineer @ Velotio (R Systems)

Pune · Apr 2021 – Present

Lead engineer on client engagements at Velotio, R Systems' product engineering arm. Each one is a different company's codebase and constraints, which in practice means arriving mid-flight, learning a domain quickly, and owning delivery rather than a ticket queue.

  • Build backend services and own the data pipelines behind Scope, the production-scheduling tool Netflix content teams work in daily, shipping into it without destabilising it
  • Took a Slack-native coaching assistant from design to production on LangChain and Vertex AI, grounded in the company's own material and cited so answers are checkable
  • Led a 12-engineer team on an event-driven go-to-market platform, adopted by 5+ teams across the business
  • Led delivery on a digital evidence platform for law enforcement: resumable uploads, duplicate detection, case workflow tasks, and API integrations with partner evidence systems
  • Built the workflow scheduler behind that platform, running 1,000+ workflows daily, and the analytics path streaming Kafka into Snowflake
  • Engineered the assistant's hard parts: context assembly under a fixed prompt budget, Slack thread reconstruction, and declining rather than improvising when retrieval came back empty
  • Ran evaluation off real production questions kept as regression fixtures instead of a synthetic set, and kept the service serverless on GCP so cost tracked usage

Client Engagements

Generative AI & LLM

  • AceUp — Slack-native coaching assistant on LangChain and Vertex AI, grounded in the organisation's own material

Media & Public Safety

  • Netflix — Backend for Scope production scheduling, ownership of its data pipelines, and a translation service for Creative Equity
  • iCrimeFighter — Digital evidence management for law enforcement: resumable uploads, duplicate detection, case workflow tasks, and partner API integrations

Enterprise Data & GTM

  • Pure Storage — Go-to-market cloud apps and APIs: the SSOT scheduler on Step Functions, and Pure1 APIs serving Salesforce and Snowflake data

Tech stack: TypeScript, Node.js, LangChain, Vertex AI, GCP Cloud Functions, AWS Step Functions, Lambda, EKS, Kafka, Snowflake

Past Experience

Co-Founder & Lead Software Engineer @ Boltpics

Pune · Feb 2019 – Apr 2021

  • Built and operated a production computer-vision pipeline processing 1M+ images/day, matching faces across event shoots in 2019, before the current wave of ML tooling
  • Grew the product to ₹1 Cr (~$120K) revenue in two years as co-founder
  • Owned the technology end to end, including the infrastructure bill and the architecture tradeoffs a compute-heavy pipeline forces on a startup budget

Tech stack: Node.js, AngularJS, Firebase, TensorFlow, AWS

Co-Founder & Software Engineer @ Busflix

Pune · Aug 2016 – Jan 2019

  • Built the Raspberry Pi software framework behind an in-bus infotainment system, and the link from a fleet of those devices back to its cloud infrastructure, in Node.js
  • Shipped over-the-air updates, hardware locking, source-code protection, and network throttling, so deployed hardware could be maintained without physically reaching it
  • Implemented the hardware interfaces the product ran on, including serial and I2C, in Node.js
  • Delivered Canstores alongside it: a multi-stakeholder retail marketplace with role-based dashboards for merchants, retailers, and buyers, plus iOS and Android clients in Ionic

Tech stack: Node.js, SQLite, AngularJS, Raspberry Pi, Linux, Ionic, AWS

Certifications

  • Claude Certified Architect – Foundations (CCA-F), Anthropic
  • AWS Certified Developer, Amazon Web Services

Education

B.E. in Electronics & Telecommunication Engineering
MIT, Pune · 2012 - 2016
Selected work
Shreyas Ponkshe

© 2026 Shreyas Ponkshe