Vaibhaw Krishna

AI Systems Engineer & Builder

I build AI-powered systems, distributed infrastructure, and products that interact with the real world — from edge AI and computer vision to spatial computing, real-time systems, and production backends.

Research → Architecture → Build → Measure → Ship

4+
Years shipping production systems
500K+
Events/day on platforms I’ve built
44%
Faster deploys after CI/CD rebuild

Selected Work

🌙

Nightwatch

AI video intelligence at the edge

Motion-gated edge pipeline that turns CCTV feeds into real events — YOLO and Gemini Vision run on-premises, so only event payloads and snapshots leave the building.

Camera→Motion Gate→YOLO→Gemini Vision→Event→Cloud
GoPythonFastAPIYOLOGemini VisionWebRTCRedis
◈

Spatial

Deterministic spatial computing infrastructure

A sensor-agnostic foundation for perception and physical-world interfaces — typed world model, replaceable device adapters, running on nothing more than a webcam and a projector.

Sensors→Perception→SpatialWorld→Interaction→Projection
PythonFastAPIWebSocketReactComputer Vision
⚡

Streamshort

Video infrastructure + creator monetization

Go backend for short-form video streaming — uploads, HLS delivery over CloudFront, subscriptions and creator payouts, deployed with a full CI/CD pipeline on AWS App Runner.

Creator→Upload→Processing→Storage→HLS→Viewer
GoPostgreSQLHLSAWSRazorpay
◆

Influenzer

Creator × brand marketplace

Clean-architecture Go service behind a B2B creator marketplace — campaigns, proposals, real-time messaging, and a Razorpay escrow/wallet system underneath.

Brand→Campaign→Creator→Proposal→Escrow→Payout
GoPostgreSQLWebSocketsS3Razorpay

Research

Edge AI for Real-Time Video Intelligence

Can motion-gated edge detection cut the cost and latency of multimodal video AI without losing accuracy? Nightwatch's pipeline runs cheap local detection first and only escalates to Gemini Vision when something actually moves — instead of sending every frame straight to a multimodal model.

Camera→Motion→YOLO→Gemini
vs.
Camera→Gemini

Measuring: Latency · Cost · Bandwidth · Accuracy

How I Build

I prefer difficult problems over fashionable technologies. I start with the problem, research the constraints, build the smallest useful system, measure it, and iterate.

Problem → Research → Architecture → Prototype → Measure → Ship

Currently Exploring

AI Agents

Reliable tool use, evaluation, memory and orchestration.

Edge AI

Running intelligence closer to physical data sources.

Spatial Computing

Sensors, perception, world models and physical interfaces.

Distributed Systems

Reliability, event-driven architecture and real-time infrastructure.

Stack

Systems
GoPythonPostgreSQLKafkaRedis
AI
Computer VisionLLMsRAGAgents
Infrastructure
AWSDockerKubernetes
Product
Next.jsReactFlutter

Building something technically difficult?

I'm interested in AI infrastructure, intelligent systems, spatial computing, distributed systems and ambitious early-stage products.

Let's talk →