Founding Engineer · Senior Software Engineer
Currently architecting a multi-tenant Model Context Protocol (MCP) platform at Clientell AI, powering AI agents for Salesforce automation across 1,000+ organizations and 2,000+ daily executions. 4.5+ years across Python, Go, FastAPI on AWS & GCP.
// about
I'm a software engineer based in Bangalore, currently the founding engineer at Clientell AI. My work lives at the seam between AI and infrastructure: building the platform that lets AI agents actually do things inside a company's tools.
In practice that's a Model Context Protocol layer across nine providers (Gmail, Outlook, Calendar, Drive, SharePoint, Teams, Dropbox, Jira, Salesforce), an LLM router that decides what each request really needs, and an async execution engine that keeps tail latency low when a thousand organizations hit it at once. The fun part isn't the demo; it's making it hold up at 2am without anyone paging me.
Before agents were the hype, I spent four years on the unglamorous things that make products work: ERP integrations that can't drop a record, event-driven pipelines, serverless jobs processing a million users, and the latency and observability work nobody notices until it's missing. That's why I treat LLM systems as production systems first: guardrails, evals, and clean failure modes, not just a clever prompt.
I like problems where correctness and speed both matter, small teams that ship, and code someone else can run without calling me. I care about the parts that don't demo well: idempotency, backpressure, retries, and the 3am pager.
Lately I'm most drawn to where AI agents meet real production constraints, cost, latency, observability, and safety. If that's what you're building, or you're scaling a backend that has to stay up, let's talk. I read every message.
// selected work
A few flagship builds: the architecture, the constraint, and the number that moved.
Architected a Model Context Protocol platform integrating 9 enterprise providers (Gmail, Outlook, Calendar, Drive, SharePoint, Teams, Dropbox, Jira) with parallel multi-source query merging, the backbone for AI agents across 1,000+ organizations.
Replaced brittle keyword routing with a hierarchical LLM classifier and a lightweight MCP agent that bypasses a 225KB unified prompt, eliminating tool-call loops and cutting connector-query context overhead by ~95%.
Migrated the agent execution layer from sync to async (Redis queues and tuned per-worker connection pools across 11 Uvicorn workers) for higher throughput and far lower tail latency.
Built a two-phase agent on GitHub MCP (repo discovery → generate-and-push) producing Salesforce Apex/LWC for org-shared repos, plus a Slack-native AI assistant that grew automation adoption and a guarded prompt-enhancement service.
Led QuickBooks, NetSuite & SOS Inventory integrations with 100% data accuracy, an event-driven sync layer (GCP Pub/Sub, Cloud Tasks, CDC), and an image service cut from 10s → 2s.
Scaled campaign processing to 1M+ users on serverless AWS Lambda + SQS, cutting runtime from 5 hours to 30 minutes, and automated AWS provisioning to save hours per client.
// career
4.5+ years across an AI-agent startup, high-growth SaaS, and enterprise platforms.
Founding engineer building the AI-agent platform for Salesforce automation, serving 1,000+ organizations and 2,000+ daily agent executions.
Owned core integrations and data infrastructure for a B2B commerce platform.
Built distributed, serverless backends for large-scale customer-engagement campaigns.
Worked on telecom-scale data analytics products and core platform services.
// toolkit
Python
Go
FastAPI
TypeScript
Node.js
Django
PostgreSQL
Redis
Kafka
Elasticsearch
GCP
Docker
Nginx
AirflowAnthropic certified: Building with the Claude API · Intro to Model Context Protocol · Claude Code in Action
// work with me
Open to a small number of engagements alongside my full-time work. If you're building AI agents or scaling a backend, or want to collaborate, let's talk.
Design and ship production LLM agents: MCP servers/connectors, routing, tool execution, evals, guardrails, and observability. From prototype to reliable production.
Event-driven architecture, async pipelines, API & microservice design, latency and cost optimization, and observability for systems that need to stay up at scale.
Hands-on technical partner for early-stage teams: architecture reviews, 0→1 builds, integrations, and helping you move fast without breaking production.
Developer-tooling and AI brands: technical content, deep-dive walkthroughs, and paid collaborations to a focused engineering audience (12K+ on LinkedIn).
// get in touch
Consulting, collaborations, or just a good engineering conversation, I read every message. Fastest is email.