I turn difficult operational problems into inspectable AI systems and shipped products. The model is one component; policy, evidence, interface design, runtime control, and human judgment are what make the system real.
| I build products around painful, high-leverage workflows: failed deployments, unsafe agent actions, operational context, and real-time decision systems. | Agent infrastructure, runtime control planes, simulation, evidence pipelines, local AI, full-stack product engineering, and operational UX. |
research → prototype → instrument → verify → ship · Bengaluru, India · building globally · open to ambitious collaborations.
|
|
REAL PRODUCTS · LIVE INTERFACES · CAPTURED 05 AUGUST 2026
|
Evidence-first failed-deployment response. Grounds itself in runtime, code, configuration, Git, and runbooks; prepares one bounded repair; verifies it; opens a human-reviewable pull request.
|
|
|
Autonomous AI runtime control plane. Connects traces, policy hits, risk scoring, intervention, and operator evidence before unsafe agent behavior reaches production.
|
Cinematic airport digital twin. Simulates 600 passengers, live disruptions, aircraft operations, and a deterministic optimizer inside an interactive 3D command surface.
|
|
Operational context control plane that keeps humans and agents working from the same configurable, verified state.
|
Local sensor-fusion garden intelligence for Arduino that stores evidence in SQLite, derives trends, detects contradictions, and recommends actions.
|
|
Procedural fantasy-anime short rendered in real time with deterministic cinematography, generated audio, narration, and film controls.
|
|
Languages · Product · Interface
Data · Infrastructure · Delivery
| AI SYSTEMS agents · evaluations · local models · guardrails |
PRODUCT interaction · full-stack · rapid prototyping |
PLATFORM observability · Kubernetes · automation |
RESEARCH algorithms · simulation · secure systems |
I contribute fixes and features upstream—not just forks. The work spans developer tooling, databases, web standards, ML systems, and agent evaluation.
|
Added a configurable right sidebar to the documentation framework. 
|
|
|
Added explicit TypeScript quality checks to the pull-request template.
|
Corrected the Python onboarding polling reference for new SDK users.
|
CURRENT UPSTREAM LANES
 Pull Shark Pull requests merged |
 Quickdraw Fast issue resolution |
Signals / year Public + private contribution activity |
PR contributions Across the last year |
Verified contribution snapshot: 05 August 2026. Live telemetry below continues to evolve.
