Founders: Brian Benedict, Mark McQuade, Jacob Solawetz
Founding: 2023
Mission: Pioneering end-to-end small, specialized, secure and scalable language models
Employees: 15 & 10% Local
Workplace: Remote
Stage & Capital Raised: Seed & $5.5M raised
Investors: Long Journey Ventures, Flybridge,
Key Customers: PIMCO, workhuman, Guild, AngelList, AngelKids
Glassdoor Rating: N/A
Valuation (estimated): $25M – $50M (assuming they sold ~10-20% of the company in the $5.5M Seed fundraise)
^ this is a useless number from MGMT Boston. There is no tangible valuation until the business is sold or goes public. Don’t forget it!
Arcee is an AI startup building Small Language Models (SLMs) so companies can deploy their own proprietary models quickly & cheaply without hallucinations or purchasing GPUs. This team is building domain specific models for enterprises & startups who don’t need to present their data in the prose of an 18th century poet.
Arcee was started by Mark McQuade & Brian Benedict who worked together as early employees at Hugging Face. There, they helped lead Hugging Face’s monetization strategy at the forefront of the Generative AI explosion. Brian & Mark talked to a LOT of enterprise & mid market companies working to figure out how to build and use AI internally. Unsurprisingly, there were gaps.
Mark went to work for another company but, after ChatGPT 3.5’s launch, he called Brian and said “this is the moment, we’ve got to build something in the AI space”. Jacob Solawetz joined to round out the Arcee co-founder trio and they were off to the races in the spring of 2023.
Brian, Mark & Jacob set out to give companies a better way to use their data to build proprietary models. Let’s be real, LLMs (large language models) are awesome at extrapolating very general knowledge bases (at least today). However, if you need LLM capabilities for a more limited set of use cases from a dataset you own, it becomes harder to customize.
Just to state the obvious, AI is booming! There are something like 67,000 AI startups building globally in 2024 (src). AI startups raised $40B+ in 2023 (src). You know this, I know this, we all know this. But Enterprise AI spending is still a minority of overall enterprise software spend so we are really *just* getting started.
The advantage Arcee brings to the market is their deep domain knowledge of Enterprise needs. Companies want new business applications that can replace manually intensive tasks or processes. Or, even better, create new revenue streams for their shareholders. In both cases, details matter. These details live in the company’s own data formatted in their internal language, culture, operating procedures, and other individualized details of the business.
If you’re not able to successfully inject your data into a LLM to get the level of performance you need, what do you do?
Arcee wants to own the model training component of small language models to solve that very problem. Their selling points are simple – give organizations a way to build, deploy, and own their proprietary models at a fraction of the cost without unpredictable hallucinations.
Model pre-training (PEFT) helps decrease compute and storage costs with comparable levels of accuracy. LoRA (low rank adaptation) techniques make sure only subsets of the neural networks are adapted during training, reducing costs too. Most importantly, Charles Goddard’s open source model merging work (mergekit) helps audit & select new open source models across the ecosystem based on what fits best for each individual customer – Mistral, Llama 3, PaLM, etc. Their evolutionary model merging software picks the base model that will be most performant based on customer data. It’s worth noting that on Hugging Face today, 75 of the top 100 models are in fact merged models.
Arcee can deploy customer models into production on a client side VPC (virtual private cloud) within a couple weeks and consistently under a month from asset gathering to deployment. As Arcee rolls out their SaaS offering in the coming months serving startups to F500 companies, all customers will need to do is upload their data, pre-train a model, merge it, and then get to work building out their own use case – all with button clicking.
Companies want more ownership, performance, and something they can speak to confidently in terms of deployment that is de-risked from the closed source LLM platform players like OpenAI. It also helps that they don’t need to spend $10M for a <$1M revenue lift.
Arcee has only been out in the market selling their product since October 2023 and has already surpassed a seven figure ARR run rate for 2024. They’re currently growing 40% quarter over quarter! Hiring will definitely increase this summer and their team of 15 will hopefully double into the end of 2024 across most functions – including GTM, Engineering, and Research.
Naturally they’ll be looking for people interested in building their career in AI and would love for many of these jobs to be in Boston where Brian Benedict calls home..
They work with small startups and enterprises and will be expanding both the customers and verticals they serve as their SaaS offering is unveiled. AWS and MongoDB are two big partners they work closely with and they most recently won MongoDB’s 2024 AI Partner of the Year award with more partnerships to be announced in the quarters ahead.
Operators to Know:
- Malikeh Ehghaghi, Applied NLP Research Engineer
- Charles Goddard, Senior Research Engineer
- Vladimir Karpukhin, AI Researcher
- Tyler Odenthal, Solutions Engineering Lead
- Mary MacCarthy, Head of Growth & Marketing
- Luke Meyers, Applied NLP & Research Engineer
- Michael Nason, Founding Software Engineer
- Gomathy Venkat, Senior Applied NLP & Research Engineer
My investigative powers continue to need work so apologies to the Arcee team if I missed any up & coming operators internally
Key Roles To Be Hired:
- More roles coming soon!
If I were interviewing here are some questions I’d ask:
- Could you share some details about the goals for 2H 2024?
- What are the biggest challenges as you navigate a crowded AI market?
- What is the long term vision for the company?
- What are the most important milestones you’ll need to achieve to be “enterprise ready” in 2024 and beyond?
We’re optimizing for readability here so to learn more about Arcee you’ll have to D.Y.O.R. I’m excited to watch this team bring more enterprises into the AI age. All innovators applaud your efforts. See you around town and the internet!