I got to sit down with Andrew Engel this week. Andrew and I go way back to the DataRobot days, where I was actually his first boss. We haven't talked properly in ages, so hearing where life has taken him was worth the wait.

The Winding Road After DataRobot

Like a lot of us who came out of DataRobot, Andrew stayed in the data science and B2B SaaS orbit for about six years. He bounced through a data engineering company trying to be the DataRobot of data engineering, then moved to Weave.ai, a consulting-style firm building customized data science solutions as a service.

The timing at Weave was interesting. Andrew joined right around the GPT-3.5 moment, and the company's CEO had spent years in traditional NLP, the kind where you spend 18 months building a custom model for a specific data type. Suddenly, he could do it with a prompt. So they pivoted, pushed into insurtech, automated underwriting, that whole world.

But Andrew hit a wall that I think a lot of people in this space have hit: they were solving problems people didn't actually need solved. That makes selling really, really hard. And the broader LLM market wasn't helping. Andrew loves large language models as tools, but he's disillusioned with the market around them. In just the last month, three different companies pitched him on LLM-based product management tools. His reaction: that's not a pain point at my stage, and I could write that prompt myself. He got a LinkedIn message from an LLM-powered SDR tool where the outreach itself was gibberish. Didn't get his job title right, didn't get his background right. A lot of noise without a signal.

The Leap to Founder

Andrew never wanted to be a founder. He said so directly. At DataRobot and everywhere after, he liked that someone else worried about the headaches while he showed up, did good work, and went home.

But something shifted. Early-stage startup culture has changed. Companies with 20 people now expect you to be available at 6 AM on a Saturday and get annoyed when you're not. And as Andrew pointed out, the payoff is asymmetric. Holding 1% at Series A, with all the liquidation preferences stacked against you, is a fundamentally different deal than what the founders get for the same hours.

When the job market seized up last year, especially for experienced data scientists, Andrew's friend came to him with an idea. The calculus was simple: what do I have to lose?

And something interesting happened once he made the jump. That 10 PM Saturday debugging session stopped feeling like a grind. When it's your thing, your creation, the same work that felt soul-crushing as an employee becomes energizing.

NYST.AI: Concussion Detection From a Phone Camera

Andrew's new company is called NYST.AI.

The concept: use a cell phone camera to capture someone performing a series of specific movements, then apply pose estimation and machine learning to detect whether that person shows signs of concussion. They're looking at how people move when they're impaired versus when they're not, and picking up on signals in that movement data.

They have a proof of concept that works on a small dataset. The challenge now is data collection, because (as Andrew put it) people don't let you go around giving them concussions to run your tests. They're working to set up design partnerships with medical institutions and navigating institutional research boards, which is new territory for a guy who spent his career in industrial engineering where nobody cared what the factory thought of the experiment.

The market is wide. Emergency room doctors are generalists, and concussions are notoriously hard to diagnose. Primary care physicians often have only superficial training on them. And then there's the massive parent market, every parent who has watched their kid take a fall and agonized over whether it's serious enough for the ER. Sports is another obvious one, given the concussion crisis across professional and amateur athletics. And Andrew lives in San Diego, surrounded by military personnel dealing with head trauma from both active conflict and routine training.

They're also tackling the edge computing challenge. Running inference on the device would sidestep a huge number of HIPAA complications. Keeping health data on the phone instead of sending it to the cloud is a meaningful privacy advantage.

As of recording, the company is stealth and is formally incorporated. They've just started receiving checks for their angel round.

What's Next

Andrew and NYST.AI are raising the angel round, signing design partners, collecting data, and building better models. The fact that it could help people, from parents worried about their kids to athletes to military service members, gives it a weight that a lot of the SaaS world just doesn't carry. If you're interested in learning more about NYST.AI or connecting with Andrew, reach out.