Why I Invested In Vellum AI & Akash Sharma

Why the space between foundation models and API's is my favorite

Talk about things that suck; the list goes; unrequited love, stepping on a Lego, and not getting into a financing round you know is going to be a winner, in that order. 

When I first heard of Vellum on Linkedin (I despise that website; it sucks the joy out of life; maybe that's why I write such off-color articles to bring some funk to the barren soulless wasteland that is LinkedIn, a site where creativity and personal expression is scorned in the name of announcing an unpaid internship) and saw the slick interface, I knew I was investing, it was clear, it was that simple. 

The very thesis of Vellum was not only model eval but dynamic model eval relative to the user's goals, not static in the way that Hugging Face had their V1 leaderboard setup. I knew there would be mounds of clunky application layer companies (if you've read my previous articles, I despise API-only companies masquerading as Deep Learning companies). Still, to my surprise, some of these companies can and do generate revenue. With that revenue comes great responsibility, and with great responsibility comes a great need for high-quality infra (this is a direct quote from Spider-Man (2002)). The thing is, Vellum was not building foundation model-level infra; no, that's a race to the bottom, a fight for crumbs for those with a passion for melting GPUs. What Vellum provided was infra for those frequently building on the APPLICATION layer! 

You see, about a year ago, I was having an existential crisis, one which, now that I think about it, has lasted my entire conscious life; regardless, I was struggling with the fundamental questions; what the fuck is investable? I'm among the few investors who get 80-plus percent of the tech, but sometimes that does not correlate to a 1:1 investment thesis. Pure play models were out of the question; application layer is for those who moved from SaaS and want to do AI, love metrics, but may not understand how wildly risky it is to invest in application layer metrics (due to transient user bases), so here I am looking a Vellum, a company that not only has a mature product but it sits between the two uninvestable areas. 

While Vellum has all of the eval, testing, and monitoring tooling, the real power comes from the space they live in; Vellum lives in the middle layer, which evaluates foundation models in the context of a specific API user's workflows and queries. This is monumental because the number of companies that build off APIs is huge. The major foundation model providers don't provide guidance on how their model works for a specific user's use case; before Vellum the user was personally doing trial and error with no infra! Seeing this problem, Vellum standardized the eval, monitoring, and testing, provided the infra to personalize a foundation model's relationship with its end users, and, in the process, Vellum opened up a huge cross-selling market.

All that to say, I knew I wanted to invest, but how? I called a friend who shall not be named because I saw we were mutuals with Akash on Linkedin and asked for an intro. He told me he had already attempted to invest, but the company said it was a no-go due to the remaining allocation. I got the intro anyway, but I was told the same thing initially, and then I started writing long and mildly technical emails about how the team could improve their product offering and GTM. One, two, maybe even three went unanswered, and then, boom, I got a meeting. Akash and Sidd were sitting in the same frame, looking at me like, who is the guy who's been spamming us but also might make a reasonable point here or there; they must have been thinking. We spoke for about 30 minutes; Akash was friendly, as brilliant as it comes on fundraising strategy, and Sidd, the king of all things dev mindshare. I could tell that with Akash's confidence, cadence, and deep love of the product and Sidd's obvious dev level of engagement, these two guys and the rest of the Vellum team would build the best product and be a billion-dollar company. So, I asked how much remained in the round and took 100% of it. 

Looking back almost a year later (I'm going to be very humble here), I got nothing wrong; these guys have shipped the hell out of their product; it is the winner in the space; anyone who disagrees does not know the product nor Vellum's metrics (what, who said that). I've loved working with Sidd and Akash, and if I were looking to go a winning startup, it's hard to argue that Vellum ain't fighting with the best of 'them' for the top spot.

Note to the reader: if your name is Akash: Akash, hook me up with more allocation. I recently bought a 2010 Accura MDX with 100K miles on it, but I'm willing to liquidate my prized MDX to own more Vellum. I'll even name the replacement bicycle "Vellum Series A."