The startup specialises in providing deployment solutions for large language models (LLMs). It offers the Takeoff Inference Server, a software platform designed to help machine learning teams deploy LLMs on various hardware systems by optimising inference and model compression. It focuses on model scalability, deployability, and infrastructure scaling features, making LLM deployments faster, cheaper, and more efficient.

In Meryem’s own words, TitanML: “helps enterprises deploy generative AI applications securely within their own environment. So instead of having to send their data off to a third party, like OpenAI, they can deploy it within their own cloud and keep it all where they are. We’ve built all the software infrastructure they need to be able to do that.”

The inspiration behind TitanML

Before starting the business, the founders could see the growth potential for AI. “At that time, we knew that AI was going to be huge,” Arik explained.

“Everyone was focusing on training and on how expensive it was to train these LLMs. We thought at the time that actually the bigger challenge wouldn’t be training the models but actually implementing them and running the models.

“We had expertise from that from my co-founders’ PhD work, in how we can make inferencing these models more efficient, and that's what we decided to go after. So, we knew AI was going to be big and we made the, at the time, controversial bet that the biggest challenge would be the implementing of these models.”

Accelerator programmes

Accelerator programmes have proved themselves to be invaluable to a startups’ success, and this has been no different in TitanML’s experience. The startup has been a part of multiple accelerator programmes: “We are a part of Grow London and London and Partners, which took us on a trade missionto the US. We’re also part of Accenture’s Fintech Innovation Lab as a lot of our clients are finance. When we very first started the company, we were in the Conception X Programme, which is designed specifically for PhD students commercialising their research.

“Out of all of them, Intel Ignite has been a really fantastic experience. Partly because of how hands on they are, and because they're backed by Intel, you have this almost unlimited depth of resource that you can bet you can pull up on, and it’s also deeptech focused. It's very different being in an accelerator for a deeptech company compared with just a regular SaaS or software company. They were some really unique things we got from Intel that we didn't get this early on.”

TitanML was part of the Intel Ignite accelerator programme as part of its Autumn 2022 cohort, and continues to fly the flag of the programme.

Speaking about the programme, Arik commented: “At the time we were in the R&D phase of what we were going to be as a business. So spending that time with Intel, and being able to get product feedback, was really valuable and helped us accelerate the R&D phase.

“But it's not just this 12-week programme. We have stayed in touch with them over the last few years and we're constantly getting feedback from Intel. So it’s definitely been a big accelerator for us.”

The challenges

“I would say the biggest challenge, but it’s also an opportunity, is how quickly the field is moving,” explained Arik. “The field has evolved enormously over the last two to three years with ChatGPT, but also the explosion of open source AI models as well.

“This has meant that we need to be building for what's going to happen rather than what exactly is happening at this particular point in time. We need to be constantly evolving and adapting, so our product has changed rather considerably.

“It’s also been a really nice opportunity because the rapidly evolving market means there's a lot of need and opportunity there as well.”

The feedback

Getting positive feedback is the lifeblood of a startup, and TitanML has had no shortage of positive feedback.

“All of the highlights are positive feedback. When clients tell us ‘we couldn't have done this without you’ or ‘you've made our lives 10 times easier’, that kind of feedback is really what we do it for.

“We’re in a really nice position where we're getting a lot of that feedback from our customers.”

The future

The future is looking bright for TitanML, with growth on the mind and new developments down the line.

“At the moment we’re really focused on client growth and finding exact product market fit,” Arik mentioned. “We have a lot of interesting product developments coming down the line, but our focus is making our customers very happy and really solving their pain points.”