Priyank Verma

Lead a pod of 3 engineers owning the team's self-hosted model stack end-to-end, from training data and post-training to evaluation and serving. Fine-tuned and distilled open LLMs to 90%+ of frontier-model quality on task evals, and serves 250k+ requests a day on a 16+ GPU NVIDIA H100 cluster.

Designed and built solo iOS apps: RezT, a sleep and HRV tracker, and Speko, a voice-first AI expense tracker, both live on the App Store. Also developing Almira, a US immigration app.

Built and launched Sentimate, an LLM-agent platform for customer sentiment analysis, rolled out to 10,000+ users across four regions. Conceived and built QuickSpecExplorer, a GraphRAG document-intelligence platform that cut technical-spec review time by 75%.

Took AcceptanceBooster (surge tipping) from A/B experiment to pan-India launch, lifting ride conversion by ~15%. Shipped Perception Offers across major cities, improving marketplace efficiency and cutting realisable burn.

Built a real-time chat system for internal teams handling student applications, plus a client-facing email service on custom domains and a daily university feed.

Built Shark.ai, a React Native incident tracker for industrial plants, a multi-dealership automotive CRM, a facilities expense and invoice app, and a chatbot-as-a-service widget.

Pursued a Bachelor of Technology in Computer Science and Engineering from JECRC University, Jaipur, from 2016 to 2020.