Engineering Leader · Founder · AI Platforms
Bengaluru, India
Over 9 years I've built products from zero to millions of users, led AI systems at Amazon scale, and run engineering orgs as both Staff Engineer and founder. I care about platforms that compound - agentic systems, RAG that holds up in production, and architecture that teams can actually ship on.
| users on products I founded | engineers led as Staff at Interface.AI | Graph RAG accuracy for banking | eng productivity via agentic product engineering |
| led first-gen AI-enabled Kindle at Amazon | uptime on voice AI interview platform | led fraud & abuse systems for stores | Gizmodo · Lifehacker · XDA |
Dobr.AI · Founder & CTO · 2024 – Present
Voice-based AI technical interviewing. Built cloud-native AWS microservices for thousands of concurrent interviews, an in-house agent orchestration framework (~1.5s latency), and cheating-detection via behavioral signals and validation agents.
Interface.AI · Staff Software Engineer · 2025 – 2026
Technical leader for a 40+ person India engineering org. Owned feasibility → POC → pilot → production (90%+ on-time, 70%+ pilot-to-production). Built Auto Product Engineering (agentic feature delivery, 5× productivity) and a Graph RAG knowledge platform for banking (98%+ accuracy).
Amazon · Tech Lead · 2020 – 2024
Led GenAI workflows for the first-generation GenAI-enabled Kindle (LLMs, agents, safety). Built fraud & abuse prevention for AWS self-checkout / Amazon Fresh. Automated receipt ground-truthing for Amazon GO (50% less associate effort, 3 days → 1 day). Led teams of 4–5 SDEs.
Blurbsh · Founding Engineer · 2019
First engineer. Mobile apps and backend routing redesign (>30% latency reduction).
Inpen · Founder · 2014 – 2017
Solo-built Android utilities to 3M+ users. Featured on Gizmodo, Lifehacker, and XDA. Recognized at IIT Madras and IIT Mandi.
Pane Browser with a soul. Open-source Chromium fork where the agent is native to your work: persona, memory, skills, and profiles on your machine. No Pane account. No Pane cloud. |
LoomStack Orchestration layer for AI-native engineering. Context, policy, and observability across the AI-driven SDLC. Early-stage exploration. |
|
Agentic systems & voice AI End-to-end agents that hold up in production, not demos. |
LLM / RAG platforms Graph RAG, evals, and pipelines tuned for real accuracy. |
|
Platform engineering Event-driven backends and systems teams can ship on. |
Fraud, trust & safety Detection and abuse systems at retail / marketplace scale. |
|
0→1 product + Staff-level strategy Founder depth with architecture, tradeoffs, and team leverage. |
|
Languages · Python · TypeScript · Java · Kotlin
AI · LLMs · Agentic systems · Voice AI · Graph RAG · Evals · RAG
Backend · Node.js · FastAPI · Spring Boot · PostgreSQL · Redis · Event-driven
Cloud · AWS · Docker · Kubernetes · Terraform
Frontend · React · Next.js
I write about AI engineering, platforms, and building in public on Substack (@vrabhi).
|
Microsoft killed Claude Code for 100,000 engineers
The token bills are the symptom. The real problem is coordination. |
|
Architecting LLM systems for scale
Selective routing and LLM–SLM collaboration for cost-efficient AI infra. |