This is a summary of the pitch presented at SPIN on April 15th.

Hi everyone, I’m Cesar Koirala from HumanLens. We provide automated AI testing and governance software for regulated industries — think of us as the pre-flight check for AI before it goes into production.

The Problem

Right now, someone in a regulated industry — credit, insurance, healthcare — has an AI making decisions for them. Are those models being tested properly? Are they being monitored properly? That’s the question all of us should be asking. In regulated industries, there’s a governance process: model owners build and serve a model, the governance team reviews it, and it gets approved. That’s how it’s supposed to work. The reality is broken. I was there — I worked in AI governance. Here’s what actually happens: data scientists serve the model, then a review process kicks in that takes at minimum three months. The triaging is manual. Risk ranking is manual. It’s a mess. And while all of that drags on, models can’t be deployed. The cost is real and financial. Companies are losing money to opportunity cost — over $80K per model by my estimation. And then there are the lawsuits. With AI adoption accelerating, it’s only going to get worse.

The Solution

Think of HumanLens as your pre-flight check for AI. Before a model goes into production, it needs to be verified — is it accurate? Is it biased? Is it compliant? We provide software testing through code. Your data scientists just call our API and run the checks without even leaving their codebase. That’s the promise.

Once testing is complete, data scientists use our API to push results to our governance platform. This is where data scientists speak the same language as lawyers and privacy officers. Think of it like aviation regulations applied to AI: structured, standardized, and auditable. A compliance report is automatically generated that your lawyers and board members can review with confidence — and now you’re ready to deploy.

Traction & Competitive Edge

We didn’t just research this — I lived it. I used to oversee thousands of models, so I’ve seen every failure mode. I’ve been a data scientist since 2006, and I deployed machine learning models in production back in 2017.

The AI governance category is being built right now — trust and safety in AI is only going to grow.

Our beachhead is regulated industry, starting with healthcare. And looking ahead, as world models and next-generation AI systems emerge, we’ll be there to ensure they’re safe.

Thank you.