Gaurav Jani, Senior Forward Deployed AI Engineer
From the first sketch to production: I design the interface, architect the system and build the AI that does the heavy lifting.
Selected work
About
Hi, I'm Gaurav. I like building software that feels simple, even when the system behind it isn't.
Who I am
I'm a developer who works across the whole product. I design the interface people touch, plan the architecture underneath it and build the AI that does the heavy lifting. Most of the time I'm doing all three on the same project, and that's exactly how I like it.
I care about the details most people never notice: how a button responds, how long a screen makes you wait, what happens when something goes wrong. Good software should feel calm, even when a lot is happening behind it.
What I've done
I've shipped AI systems that read, check and route documents in production, at a scale of hundreds of thousands a year. I've also built a customer-facing portal on my own, from the first design to deployment, support and the late-night fixes that come after launch.
Before that, I spent close to two years building front ends: a document workflow for notaries, and a Chrome extension with real-time previews, drag-and-drop and map-based tagging. I graduated in Computer Science in 2025.
What I'm working on
Making LLM systems dependable enough to trust with real money. That means structured outputs, deterministic checks around every model call, and prompts that can be changed safely in production without a new deploy.
I'm also building assistants that can actually do things rather than just answer questions, with function calling, voice input and memory that carries across a conversation.
What I'm exploring
How to tell whether an AI system is right when nobody has labelled the data, and how agents should hand work back to a person when they're unsure, without becoming a nuisance.
And the interface side of AI: how to show a model's confidence, how to let people correct it, and how to make all of that feel natural instead of technical.
Models are most confident right before they're wrong. Anything that touches money gets a deterministic check.
Off the clock: coffee, good typography and websites that do something unexpected.
Say hi!
My inbox is open.
Read something interesting, building something fun, or just want to swap notes? I'm always up for a good conversation.
Download resume