OVERVIEW
Principal Software Architect and Founding Engineer specializing in the architecture of high-scale, AI-integrated real-time platforms. Instrumental in spearheading the Vooz ecosystem from initial conception to a global social infrastructure serving over 500k monthly active users. Expert at navigating the intersection of complex frontend orchestration, distributed backend systems, and high-performance AI inference.
SKILLS
FRONTEND
Next.js 16, TypeScript, Redux Toolkit, WebRTC, Framer Motion
AI_INFRA
Triton Inference Server, ONNX Runtime, Dynamic Batching, Model Quantization
DEVOPS
Azure AKS, ArgoCD (GitOps), Azure Front Door, OpenTelemetry
WEB3
Node.js, Solana (web3.js), SPL Tokens, Helius Webhooks
REAL_TIME
SignalR Hub Sync, Socket.io, P2P Data Channels, MediaStream
DATA
Memgraph, PostgreSQL, Redis, Cache API, Persistence Layer
EXPERIENCE
VOOZ INC
Principal Software Architect (Founding Engineer)
- Engineered multi-tenant Next.js environment utilizing Parallel and Intercepting Routes, scaling to 500k+ MAU with optimized SSR delivery.
- Designed and maintained Vooz Video Coordinator for WebRTC handshakes and Media Recovery Logic.
- Spearheaded Triton Inference cluster with Dynamic Batching, achieving 1280+ RPS with full OTel tracing.
- Architected Solana payment infrastructure via Node.js microservice using Helius Webhooks for atomic transactions.
- Orchestrated multi-cloud deployments secured by ArgoCD and Azure Front Door with cookie-based edge routing.
- Established zero-latency SignalR Hub Sync protocol to recover state from abnormal disconnections (Code 1006).
- Engineered browser-local inference using ONNX Runtime (WebGPU) and Cache API for real-time content filtering.
AIMICA LTD
Software Developer (Lead Role)
- Led a team of 3 developers building AI-powered mobile apps using React Native (Expo).
- Designed and implemented a scalable RAG-based knowledge system integrating open-source LLMs.
INFOEDGE
Mobile Developer
- Optimized critical React Native flows for high-traffic apps, resolving memory leaks on low-end devices.
PROJECTS
Medical Paper NLP Categoriser
Deep learning model using TensorFlow on PubMed datasets. Achieved 90%+ accuracy in classifying medical abstracts.
Secure Video Conferencing Suite
Cross-platform video app featuring end-to-end encryption and real-time signaling via Firestore and WebRTC.
EDUCATION
B.Tech in Computer Science
Dr. A.P.J. Abdul Kalam Technical University