AI Systems Architect
Technical Architect with 8+ years building large-scale backend, cloud, and AI platforms. I joined PhroneticAI as a 0→1 founding engineer and now hold final architecture and production sign-off across the platform, while leading the engineering team through design reviews, hiring, and delivery. I stay hands-on: the agent gateway, payments platform, observability stack, and cloud estate are systems I designed and still operate.
Technical Leadership
I lead through architecture, judgment, and the people around me. The systems on this page were designed by me, and the team that runs them was hired, mentored, and reviewed by me too.
Architecture Ownership
Final architecture and production sign-off across the platform. I write the high-level and low-level specs before a line ships, review every design proposal, and own the trade-offs for each new service: the agent gateway, runtime, billing, RAG, and infrastructure were all designed this way.
Force Multiplier
Lead a core team of five engineers and mentor beyond it: structured 1:1s, design reviews, 100+ technical interviews with final go/no-go, and sign-off on two external contract firms. The aim is engineers who own outcomes and can defend their designs, not engineers who close tickets.
Build-vs-Buy Judgment
Evaluated Killbill, Lago, and OpenMeter and built billing in-house for control over a usage-based model. Moved agents from dedicated deployments to a shared stateless pool and cut cloud spend by over 85%. Self-hosted the observability stack instead of paying per seat. Every call is made on unit economics, not habit.
Production & Security Ownership
Sole owner of the AWS and GCP estate: Terraform from org bootstrap, IAM/SSO, secrets lifecycle, on-call at any hour. Closed a payment-tampering vector with JWT-signed requests and built LLM-era controls: prompt-injection hardening, sandboxed tool credentials, and code-execution allowlists.
Impact & Achievements
Years Experience
backend, cloud, and AI platforms
Latency Cut
agent time-to-first-token: ~5s to ~200ms
Cloud Spend Cut
shared stateless runtime pool, Terraform-managed
Products at Scale
GeM and JioMart catalogs, thousands of searches/sec
Technical Interviews
SDE to AI Engineer, final go/no-go on panel
Case Studies
Systems I architected, built, and operated in production. Each one covers the problem, the technical decisions, my role, and what it changed for the business.
AI Agent Platform & Gateway
Architected the platform foundations at PhroneticAI from a blank slate: an OpenAI-SDK-compatible agent gateway, a shared stateless agent runtime, real-time usage metering and billing, multi-tenant RAG, and the observability that keeps it all honest.
PayCentral
India's first agentic payment platform: an autonomous payments layer built on Google's AP2 protocol that lets AI agents discover merchants, negotiate checkout, and complete payments through any connected payment MCP, all inside a single conversational interface.
Video Intelligence Pipeline
Built the backend, alerting, and streaming infrastructure for a real-time hospital and industrial CCTV monitoring platform: an in-house WebRTC transport, a fall/SOS/fire detection pipeline with VLM-based false-positive filtering, and an edge appliance that turns any site's cameras into cloud streams.
Marketplace Catalog & Search Platforms
Owned catalog management and search for GeM, India's national procurement marketplace, then the seller and catalog platform behind Reliance JioMart. Together: 5M+ products, thousands of search requests per second, and a sync layer feeding 7-8 downstream systems.
Cloud Platform, Cost & Observability
Sole owner of the company's AWS and GCP estate: Terraform from org bootstrap, IAM/SSO and network design, secrets lifecycle, and a self-hosted observability stack. Led two cost drives that cut monthly cloud spend by over 85%.
Professional Experience
A snapshot of my engineering journey
Technical Architect
PhroneticAI0→1 founding engineer · Tech Lead → Technical Architect (Apr 2026) · leads the engineering team
- Final architecture and production sign-off across the platform; write the high- and low-level specs before implementation begins.
- Designed and built the agent gateway, payments and metering platform, RAG service, and observability stack from zero to production.
- Lead a core team of five engineers: design reviews, 1:1s, hiring (100+ interviews), and sign-off on two external contract firms.
- Sole owner of the AWS/GCP estate; cut monthly cloud spend by over 85% and agent response latency by 25x.
Software Engineer → Senior Software Engineer
Infibeam Avenues- GeM: owned catalog management and search for India's national procurement marketplace (3M+ products, thousands of searches/sec); ~1,000 shipped work items across 5 Rails apps.
- JioMart: ranked #1-2 committer; owned the sync layer keeping 2M products, prices, and inventory consistent across 7-8 downstream systems.
- Led the Rails 7 / Ruby 3 platform upgrade and built scale-to-zero pricing job pods on Kubernetes; cut search response times by 75%.
- Mentored developers and led code reviews across teams.
Technical Expertise
~/now
What I'm heads-down on right now.
- $Agentic AI core runtime internals
- $Agentic commerce protocols (AP2, UCP) and agent-initiated payments
- $Latency and fault tolerance in distributed systems
- $Scaling technical judgment through others: design docs, reviews, mentorship
- $Connection-aware autoscaler for async Python servers
Traditional autoscaling scales on CPU/memory, but async frameworks like FastAPI bottleneck on open connections first. Building an open-source autoscaler that scales on live connection load instead.
- $Secure sandboxing for multi-tenant Python execution
An open-source sandbox for running untrusted Python code safely in multi-tenant environments.
- $Badminton
- $Chasing new places — globetrotter at heart
- $Real Madrid, through wins and heartbreaks
- $Long drives, no particular destination
- $Debugging — genuinely, even outside work
- $Running 3 aquariums in production — 100% uptime, no on-call needed