I build AI products that ship - from zero to one, and to scale.

Bolto is the AI-powered hiring and HR stack for engineering teams - sourcing, AI vetting, scheduling, payroll, and compliance in one place.

• Building AI-driven engineer sourcing & vetting pipelines that ingest signal from public code and match candidates to companies at scale.
• Owning full-stack slices across sourcing, technical assessment, scheduling, and the recruiter experience.
• Bringing the agent-orchestration and LLM-routing playbook from Spine into recruiting workflows.

Spine ships an AI research platform whose deep-research engine ranked #1 on Google DeepMind's DeepSearchQA benchmark. I owned the orchestration engine, model router, and large parts of the canvas runtime.

• Architected the backend orchestration engine for Spine Swarm - 1–10 specialized agents per task, 500+ concurrent, parallel non-sequential execution at production scale.
• Built the intelligent model router - 400+ LLMs selected dynamically per step, optimizing for quality, latency, and cost. Cumulatively processed 106k+ intelligence nodes.
• Powered 6,400+ active canvas environments with collaborative agent execution, artifact persistence, branching, and full inspectability.
• Shipped APIs that produce client-ready outputs - decks, PDFs, Excel models with formulas - and the React canvas frontend.
• Owned multi-cloud DevOps (AWS / GCP / Azure) with zero-downtime blue–green deploys; took Spine 0→1 to 10k+ users · 151 paying · 2.5% conversion.

Founding-member SDE for the company behind Padhai.ai (UPSC, 175k+ users), EverTutor (K-12 + GRE), and a portfolio of AI tutoring products. Owned backend, infra, and reliability across the entire stack.

• Owned the AI webserver behind Padhai.ai - comprehensive refactor and error-handling overhaul, with major stability gains under peak exam-season load.
• Owned EverTutor's backend & infra end-to-end - integrated Stripe for international + domestic payments and unlocked the entire revenue path.
• Led migration from PM2 processes to a containerized ecosystem; designed blue–green deploys with Docker for zero-downtime updates.
• Built a 10,000-concurrent WebSocket microservice for real-time AI interactions, plus a separate CPU-intensive recordkeeping service to keep latency low.
• Architected Datadog APM + centralized logging; shipped a Redis sorted-set leaderboard and a React Native markdown + LaTeX renderer for every tutor response.
• Managed containerized microservices across AWS / GCP / Azure; led and mentored a team of interns.

Shipped React Native + Redux app features and Django APIs/websockets for AI chat; built a Redis cache leaderboard, an FCM notification panel (bulk send, CSV upload, Celery delivery), a deeplink-driven sharing pipeline with dynamic image generation, and a D3.js chat-graph visualization for conversation inspection.

Built mission-critical POS flows in Next.js + Node used in production across stores. Shopify order sync via APIs/webhooks, Razorpay payments, SearchTap reindexing. Discovered and fixed an indefinite-OTP verification flaw, hardening the mobile auth flow.

Built an Incident Response GUI from scratch using Volatility, Log2Timeline, and RegRipper for forensic timelines. Mapped adversary tactics with MITRE ATT&CK CASCADE, built an ETL log parser for Windows pktmon/tracerpt, and reverse-engineered WannaCry on Remnux to identify its kill-switch.