Shubham Sharma
What I do
I design and build the backend systems that power AI-driven products.
AI-native systems
Agent architectures, LLM orchestration, vector search. In production, not prototypes.
Security-first
From cybersecurity at TCS to PCI-compliant fintech. I design systems with security as architecture.
Production-grade
35+ microservices, CI/CD pipelines, observability stacks. Built for reliability, not demos.
Scalable from day one
Distributed architectures, Kubernetes, Kafka pipelines. Designed for scale, not rewritten for it.
An autonomous AI business manager for the creator economy. Fluencity handles the business side of being a creator: deal sourcing, negotiation, contracts, payments. So they can focus on creating. Operating natively on WhatsApp, email, and voice.
Background
Cybersecurity → Fintech → AI platforms
Started in cybersecurity at TCS. Identity management, endpoint protection, SOAR automation. Moved to Converj where I helped build a PCI-compliant B2B payments platform: 35+ microservices in Go, AI-powered payment routing, and a 30-person engineering team. Now applying all of it to build intelligent, production-grade infrastructure at Fluencity.
Recent
From LinkedIn
Structure > smarter models
One mistake I see a lot while building AI agents: People try to make the LLM "smarter". In reality, most problems come from not having enough structure around it...
On agent frameworks vs building your own
Most "advanced agent architecture" patterns (planning layers, tool execution loops, guardrails, memory, retries) weren't new problems for us. Real product requirements forced us to build them...
Why creators miss deals
Most creators don't miss deals because of talent. They miss them because of noise. In a recent conversation, a creator showed us their inbox: 40+ brand emails, half buried...
Fluencity Public Beta launch
18 Hours. 1 Simple Ad. Hundreds of Requests! We quietly flipped the switch on the Fluencity Public Beta. We didn't do a massive PR blast, we just ran one simple ad...