Abhibhaw Asthana
Product manager who builds. I find where people wait, decide what to build and what to leave out, and ship the first version with AI agents. Currently leading product at Devtron. Products in enterprise B2B, AI agents and health.
Experience
Open-source platform that companies use to deploy and run software on Kubernetes. I lead product here, across 3 products.
AI Control Plane · build, ship and govern AI agents securely
- Agents in production: Companies had no safe way to deploy AI agents or control what they do. I defined one control plane to build, deploy and govern agents. In private beta with 3 design partners.
GPU Platform · self-service GPUs for AI teams
- Faster GPU access: AI teams waited days for GPUs and weeks to ship a model, through infra tickets. I defined self-service GPU workspaces, training jobs and one-click model deploys, cutting GPU tickets by 80%+ and making developers fully independent.
- Better GPU use: Costly GPUs sat idle or stayed locked to one team. I added team quotas, usage policies and GPU sharing (MIG, time-slicing), improving GPU utilization and allocation by 23% on average.
Devtron Platform · the core Kubernetes deployment product
- Trial conversion: Trial users left before they saw value; users who deployed within 20 minutes stayed 3x more. I rebuilt the trial around a guided first deploy: 4x inbound leads, 33% faster paid conversion.
- Enterprise approvals: 15+ interviews showed regulated enterprises could not adopt us without sign-off before production changes. I shipped multi-level approvals and deploy policies, which became a key reason they chose Devtron.
- Atlas, AI debugging assistant: Users with failed deployments had to read raw logs and often opened support tickets. I shipped Atlas, which finds the root cause and suggests a fix: root-cause time fell from hours to about a minute, and time to recovery from hours to minutes.
- Self-service ops portal: Engineers waited 14 days for a small DevOps team to run database queries and server changes. I built a self-service portal for database, Kafka, GitHub, Redash and API gateway requests, with manager approval and a Jira audit trail. It handled 20,000+ requests and cut the wait to 17 minutes.
- Safe database changes: Changing a live MySQL table could lock it and cause downtime. I added zero-downtime table changes to the portal, with a data check before switch-over and a second sign-off from DevOps.
- Software for small businesses: Small business owners needed software but had no tech team. I turned their needs into scoped products like payment dashboards and online stores, and shipped 37 apps in 18 months with a team of 5.
Projects · Independent, in development
- Sakhi: Older Indians with diabetes, BP and thyroid miss medicines unless family reminds them. I am building an AI nurse that calls them on WhatsApp in Hindi and briefs family and doctors. Its safety checks pass all 52 red-team cases.
- Dhanvantari: Indian doctors lack a trusted clinical Q&A tool that covers Indian guidelines and medicines. I am building one where every claim links to a source, and a release is blocked if citation accuracy drops.
- Peira: AI agents keep failing in production, and teams find and fix each failure by hand. I am building Peira, which finds repeated failures in an agent's traces and proposes a code fix. It tests each fix by rebuilding the apps the agent used from its traces and replaying its conversations. A person approves every fix.
Education
Skills
Product: customer discovery and interviews, specs and PRDs, prioritization, funnel and activation metrics (PostHog), enterprise governance (RBAC, approvals, audit)
AI: AI agents and their governance, evals and replay testing, RAG with citations, voice AI, LLM safety and red teaming, GPU platforms (quotas, MIG, time-slicing), model serving (KServe, vLLM, Triton)
Technical: Kubernetes, Docker, Go, Python, TypeScript, React, PostgreSQL; building prototypes with AI coding agents