November 22, 2024 Company Building

Delphina Blueprints: Freeing Data Scientists to (Finally) Focus on the Deep Work

BuilderOps Blueprints is a newsletter on company building foundations for early-stage startups by Costanoa Ventures, a VC firm that backs builders across data, dev, and fintech. Throughout this series, Costanoa’s BuilderOps team interviews founders and startup leaders, showcasing their superpowers and learnings on all things company building.


For our latest edition of BuilderOps Blueprints, we have a twofer for you in the form of a sit-down with Delphina’s two remarkable co-founders: Jeremy Hermann and Duncan Gilchrist. Delphina harnesses generative AI to automate foundational data science tasks, uncovering and cleaning relevant data to ultimately transform noise to meaningful signal. If you’re interested in working at the edge of machine learning and AI, check out open roles here

Let’s start with something fun – the name Delphina has obvious connections to the Oracle of Delphi and predicting the future. Are you mythology nerds?

Jeremy: We’re both fond of history and so loved that connection. But I actually first saw the name itself when I was coding to electronic music on Spotify – Delphina was a song by Frameworks! There’s also a light link to a favorite spot in San Francisco: Pizzeria Delfina.  From there, it fell into place. Delphina felt memorable and easy to say – and the domain name was available.  

Going deeper for a moment: what core problem are you trying to solve for customers?

Duncan: We want Delphina to move data scientists to a working environment with vastly less toil and more opportunity to use their real skills. Today they spend 40 percent of their time on preparing data – and just 10 percent on the really deep work that truly is differentiating. Imagine what companies – and what the world – would be like if they could use their talents in an unrestrained way.

Jeremy: This is a thread I’d been working on since Uber: How do we automate the painful parts of data science work? Data scientists are exceptionally smart, highly-trained people. They shouldn’t be spending the majority of their time working on tedious, mundane things! Let them focus on creative ways of working that are actually worthy of their intellects. Large language models are finally making this possible in a way that it simply wasn’t before.

Pulling on that thread, what’s the best way to validate customer pain points? We know you did a tremendous amount of work to feel confident in Delphina’s use cases.

Duncan: Well, we are data scientists so we know what’s frustrating to us! But we did a relentless amount of customer discovery – and that’s key for anyone to do. Early on we did over 100 calls with data scientist teams, probing into what value they are adding to their businesses right now versus what value they could and want to be bringing. There was a big delta in between. 

Jeremy: These calls told us that most of a data scientist’s time is spent preparing the data so it can work for them – and that most existing tools don’t help with that workflow. Other solutions are more oriented toward visualizing the data, not working with it upstream where it’s super labor-intensive. This means good data science is slow for the best companies and prohibitive for everyone else.

You might be our newsletter’s first double interview so let’s ask the obvious. What makes you click as co-founders? And what advice do you have for other entrepreneurs to find the perfect match?

Jeremy: One of the hardest things about founding a company is it feels like pushing a rock up the hill. A lot of it is keeping your energy and mood up so you have to find a partner who balances you. I often find on days when I’m down, Duncan is up and vice versa. We’re both very technical so that overlap helps us collaborate.

Duncan: We have different backgrounds but similar personalities. We’re both even-keeled so we don’t get enraged at things. We work really well on that EQ level, and it creates the magic. We both go very deep into topics – but in different areas. And we both have families with kids too, so we’re building two important things personally and professionally.

How do you manage kids and being founders? That’s a concern for many founders – and it’s a hard balancing act.

Jeremy: On a company level, we’re very clear about supporting family. That flexibility is baked into our ethos of being a culture where we prioritize outputs, not necessarily time spent. It enables us to hire other like minded experienced engineers and scientists. Personally, I get up early in the mornings before the kids are awake and that’s when I get focus work done. 

Duncan: I’m finding it very rewarding to balance time with the kids and my responsibilities here. They are three and five so when I get home, what I do there is very different from what I do here for a few hours. It’s a really nice balance to life.

Say you’re talking to a founder who’s just launched a company. What’s your advice for staying sane?

Jeremy: Exercise. We both make time for it. I run or bike outside.

Duncan: Invest actively in your own learning. You can only learn so fast through your own experiences and the best way to supplement is with life lessons from others. That’s why we both make time to read.

Whose stories have jumped out at you recently?

Duncan: My favorite book of all time is Into Thin Air. It’s an incredible non-fiction story of an Everest ascent – and all of the hairy, life-and-death challenges that get in the way as the air gets thin and judgment falls apart. It makes you reflect on the nature of humanity, and how we make decisions when really stretched. 

Jeremy: I recently read biographies of Elon Musk and Steve Jobs – both not life-changing books but really helpful reminders that most founders don’t encounter that smooth, easy, “up and to the right” growth path. The common thread was around incredible perseverance and the impact that profound commitment can have on your path to success.

Where do you want Delphina to be in the next five years?

Jeremy: Widely deployed and widely used! We want to get to the point where we’re well known because our stuff works and matters.

Duncan: Helping data scientists spend just 10 percent of their time on mundane tasks – and driving that number closer to zero.

You just launched a podcast – talk a bit about why? 

Jeremy: We’ve talked to hundreds of data science teams about where they learn – and we’ve found there’s a dearth of high quality practical content that spans the gap between papers and blogs. We have a pretty deep network of data science leaders, and we were brainstorming with a podcaster friend of – the result is now High Signal.

Duncan: Our launch episodes are with Michael I. Jordan (the “Michael Jordan of ML”), Andrew Gelman, and Chiara Farronato of Harvard Business School. And we’ve got an incredible list of speakers lined up next. Check it out! 

Author

0 Shares

Written by

Link
author

Director of Marketing & Platform

Taylor Bernal