At TSVC, we recently launched our Deep-Tech Founder Podcast - On the Edge, where we talk to tech entrepreneurs about the challenges they face and the solutions they build. Our first guest is Bertrand Irissou, CEO and Co-founder of Brainome, a company that offers the first measure-before-build tool for machine learning.
Modern machine learning, despite its seemingly mysterious nature and high complexity, has a fundamental problem. Instead of intelligent, well-conceived strategies for solving problems, too often solutions of brute force are applied, resulting in flawed outcomes. Massive computational power is thrown at large sets of data in the hopes of finding patterns. However, with a little bit of preparation, this approach can be modified, leading to more ‘intelligent’ machine learning models.
Gerald Friedland, Bertrand’s Co-Founder, is a long time researcher at UC Berkeley. He realized machine learning was very anecdotal and overly reliant on guess-and-check methods. In implementing machine learning, engineers would often automatically construct and test a large number of models each with slightly different parameters to inch closer to a decent end result.
But this process simply wasn’t scientific.
Bertrand compares this conundrum to building bridges. In the real world, a bridge is constructed using sound engineering, organisation, and analysis. Engineers methodically plan which parameters a bridge needs to fulfill to hold up. In machine learning, the current industry practice is hit-and-miss. 100 bridges are built, 100 trucks are driven over, and if luck prevails, one bridge won’t collapse. This process is neither efficient nor reproducible.
Brainome solves this challenge by taking measurements before engineers dive into machine learning. The company implements rules and design specifications to streamline the entire process for engineering teams. Brainome’s product is a data compiler that helps people go quickly from classification data sets to a working model, solving supervised machine learning problems in a repeatable way.
In our conversation, Bertrand espouses that founding a company involves the same anecdotal unpredictability that machine learning did before Brainome. As a dedicated entrepreneur with experience as a private angel investor in robotics, technology and medical device companies, Bertrand offers his advice for creating well-rounded founding teams, pitching investors, and establishing beneficial partnerships.
The most fundamental element for the success of any company is its founding team. With the right mix of personalities, strengths, and networks, the path ahead is much easier to navigate. Bertrand emphasises: “Have technical grounding but don’t forget the other aspects of company building. Specifically, ensure you have team members who are well networked in your space and can easily connect with potential customers and investors.”
Bertrand very much highlights the importance of complementary networks. Knowing the right people, be it mentors, investors, or potential employees, is incredibly important for a startup in its early stages. “Startups must move quickly with limited capital, so the faster you can get to your target audience the better”, says Bertrand.
He adds: “Beyond skill sets, make sure you hire people who respect and trust each other. Building a startup is incredibly challenging, but if you have people who believe in each other, you can have productive disagreements and arrive at sound decisions that everyone ultimately buys into.”
Ultimately it is the founding partnerships that dictate the success of a startup.
Most young founders dread pitching to investors. It’s a high-stress, high-stakes situation with an uncertain outcome. Fast heartbeat, sweaty hands, blurry vision - most people know these symptoms all too well. While experience definitely helps to mitigate the stress level, there are some more tips worth paying attention to.
Bertrand advises: “Meet investors where they are. Come up with something they relate to.” For example, with Brainome, Bertrand often starts by noting that machines learn just like humans do. Depending on the technical complexity of your product, it is important to adapt the communication to your audience. “Before going into a pitch meeting, figure out what technical comfort level the investor has and tailor your pitch based on that”, he says.
One important point he emphasises is to understand that you cannot convince everyone. Bertrand says: “Most investors will say no, and you can usually tell how a meeting will go by 10-15 minutes in.” However, the good news is: you don’t need to convince every investor. You just need to convince enough of them. “Even a few yes’s in a sea of no’s is enough,” he adds.
While fundraising is inevitably stressful, it’s also a great opportunity for founders to develop a deeper understanding of their business or product. Investors tend to drill into the assumptions behind a business and its monetization model. Of course, founders need to be thoroughly prepared for these questions, but at the same time this probing allows entrepreneurs to think deeply about their assumptions. “Even beyond fundraising, reevaluate your assumptions continuously. Assumptions form the foundation for your business decision making, so it’s crucial to ensure they are sound and up to date”, Bertrand says.
Founding a startup results in countless challenges, and more often than not in sleepless nights. While building a great product is already hard, getting other people to accept it is often even more challenging. That is why it is crucial for any company to have a strong sales and marketing team. Bertrand says: “Even the best products don’t sell themselves in the early days. Even the best introductions from your investors and network are just introductions. You and your team still have to do the hard work of selling.”
In the early days of a company, founders must look out for first adopters. First adopters are customers that are characterized by a greater tolerance of risk, preferring innovation over tried-and-tested products. But they can equally be people who simply don’t have a choice but to go with your product. Bertrand remarks: “If given a choice, people will rarely take a chance on a seemingly risky, early-stage startup. Instead, they would go with the solutions built by a large tech company. So find the people with the hair-on-fire problem that only your technology can address.”
To close, Bertrand shares what he wishes he knew starting out.
“You only get one chance to make a good first impression. You have to win the hearts before you win the brains. Make sure your first out of the box experience is amazing. Even if the technology works, you need to ensure the user interface is spectacular. Be open minded in your thinking. A plan is ultimately just a plan, and with startups, plans inevitably evolve all the time. Bake flexibility into everything you do.”
Listen to the full podcast with Bertrand here:
To learn more about Brainome and improve the learnability of your data, visit https://www.brainome.ai/