# Dheer Vora

> Product leader in Bengaluru with 13 years of building for how India actually works: payments at Paytm, food ordering at Rebel Foods, gold loans at Rupeek and workforce software at BetterPlace, after three years in manufacturing at Schneider Electric. On a deliberate sabbatical since August 2024, building hands-on with AI. Open to a full-time leadership role, a consulting engagement, or advising a team that is building with AI.

This file mirrors https://dheervora.com/ in plain text. Every figure here is also on that page.

## Contact

- Email: dheervora@gmail.com
- LinkedIn: https://www.linkedin.com/in/dheervora/

## Career

- Sabbatical, independent, Bengaluru (2024 to now): deep work on the modern AI stack (LLM architectures, agent harnesses, context engineering, orchestration, evaluation pipelines), mapping the vertical AI landscape, and building. 7 products and tools built, 4 of them public on GitHub.
- BetterPlace, Director of Product Management, Bengaluru (2023 to 2024): led seven PMs on a B2B HRMS platform for hiring, verifying, onboarding and managing blue and grey collar workers; owned product, go-to-market and pricing; $10M ARR reached; +80% sales pipeline growth; -80% operational cost via fine-tuned LLMs and ML; -60% customisation cost via a low-code builder. Led MyRocket, a volume-hiring job board that cut cost and time to hire by up to half. Helped OkayGo, MyRobin and Troopers, the group's companies in India, Indonesia and Malaysia, shape their strategy and roadmaps.
- Rupeek, Senior Product Manager, Bengaluru (2021 to 2023): pioneered the Gold Teller Machine, an industry-first hardware and software product for self-serve gold loans (under 10 minutes per loan, NPS 65+, over a thousand loans processed); designed the lending stack behind bank and NBFC partnerships (-50% lender onboarding time, -30% loan disbursal time); launched a hybrid secured and unsecured credit line.
- Rebel Foods, Senior Product Manager, Mumbai (2019 to 2021): launched EatSure, multi-brand ordering from Rebel's cloud-kitchen brands in a single order; 20% of revenue moved to own channels; +3pp open-to-order conversion; ML-driven recommendations and quick-cart checkout.
- Paytm, Senior Manager, Product Led Growth, Noida (2016 to 2019): self-serve merchant onboarding, gamified multilingual activation campaigns and real-time incentives targeted by auto-segmentation, tools for a national field sales force; 6x growth in QR merchant payments; +33% field sales productivity.
- Schneider Electric, Planning Engineer, Nashik (2011 to 2014): supply chain and production planning; +49% production output; +15% on-time deliveries.

## Education

- PGDM, S. P. Jain Institute of Management and Research (SPJIMR), Mumbai, 2014 to 2016
- B.Tech, National Institute of Technology (NIT Trichy), Tiruchirappalli, 2007 to 2011

## Lab (public on GitHub)

- [Claro](https://github.com/dheervora/claro_mvp): in-app onboarding assistant; an AI cursor guides users by voice or text and offers help when it detects confusion. Next.js, OpenAI, Deepgram, Pinecone.
- [SenseIt](https://github.com/dheervora/senseit): macOS dictation that reads the active app's context and turns Hinglish into clean professional English. Python, macOS accessibility APIs, LLMs.
- [Ledger](https://github.com/dheervora/expense-drive-bot): log an expense by sending "30 tea" on Telegram; stored AES-256 encrypted in the user's own Google Drive, no database. Next.js, Vercel, Telegram, Google Drive, Gemini.
- [CinePrompt Studio](https://github.com/dheervora/CinePrompt-Studio): learn 48 photography and film terms, then use them to prompt image and video models. React, TypeScript, Gemini.

Private projects, demo on request: Groove Atlas (29 interactive lessons in rhythm, harmony and style), Out of the Cave (the Philosophize This! podcast mapped into 1,030 linked ideas and 250 thinkers), Chord Deck (a phone-sized chord reader and editor over a Google Sheet).

## Views on AI products

- The moat is the harness. Holding the model fixed, harness changes alone moved LangChain's coding agent on Terminal Bench 2.0 from 52.8% to 66.5%. The general-purpose layer belongs to the labs; what lasts is the domain layer of tools, checks and permissions.
- Clever plumbing has a short half-life, and a long tail of cost. Workarounds for today's model limits go stale with each model generation but keep costing tokens, latency and re-testing. Treat them as debt and audit them at every model release.
- Enterprises will own their context. Vendors will rent it. Decision context draws on every workflow and needs governance an enterprise will not hand over, so builders should own hard-to-synthesise data and be the best place to act on context the customer keeps.