When Nicolas introduces himself, you immediately sense the structured, calm, pragmatic profile:

"I've always loved understanding how things work"

He starts at the beginning:

"I'm an engineer by training. I did a prep program then engineering school, and that's where I really discovered computer science… but always through a data lens. I had a very math-heavy background at the start."

That first seed leads him to Boston, where he does a master's in machine learning:

"That was about ten years ago. AI was already a very hot topic in academia, even if it was nowhere near as mainstream as it is today."

After Boston, he moves to San Francisco to join Salesforce:

"I joined a small team that came from a startup acquisition. We were heavily specialized in AI and data science. It was a very challenging environment. The team grew fast, we were learning every day."

He stays a year and a half before returning to France.

Two formative experiences await him: AnotherBrain, then Iktos.

"AnotherBrain was very research-heavy. The goal was to miniaturize embedded AI inspired by how the brain works to automate manual processes in factories. Super interesting intellectually, but too far from real-world application for me."

He then moves on to Iktos, a deeptech startup in healthcare:

"There, I loved it. We were developing generative models to identify candidate molecules for future drugs. It was the first time I tackled healthcare at scale. A perfect mix of research and development, because there was a real product used by real chemists."

But even there, something is missing.

"Biology, chemistry… it's fascinating, but extremely hard to master. These are fields where experts have ten or twelve years of experience. I wanted to come back to a topic where I could understand things faster, be more useful. And I wanted to get closer to a more tangible impact, something whose human stakes you immediately grasp."

"Klineo came at the right time"

That's when he reconnects with a former classmate: Thomas, who was launching a healthcare project with Arnaud, a medical oncologist.

"They were already talking about the problem of clinical trials. It's a nightmare for doctors and patients to access the right info at the right time. They had the business and domain vision, the tech vision was missing. The alignment was perfect."

He sums it up simply:

"Today, an oncologist can spend an hour just checking whether a patient is eligible for a single clinical trial. There are thousands of them. It's impossible to do that for every patient."

Klineo offers a platform that can:

• search for a clinical trial for a patient in less than a minute
• aggregate clinical trials from different sources
• standardize heterogeneous medical data
• accelerate the entire referral workflow into a clinical trial
• secure the exchange of a patient's medical data

"In France, more than two thirds of oncologists use Klineo — penetration is very strong."

The startup recently expanded across Europe, with Belgium, Germany, and Italy.

"Our biggest challenges are data, data, and data"

When asked about the technical challenges, he doesn't reach for fancy phrasing:

"Clinical trials are unstructured data, published on multiple sites, never the same way, updated at different frequencies, with often very specialized medical content. It's hellish."

He elaborates:

"One source will have the update date. Another will have inclusion criteria. Another the interventions. Another the investigation centers… You have to reconcile everything. It's huge."

Then comes the search engine, which he describes as the heart of the system.

"At first, it could take five to ten seconds to return a result. Now we're at less than a second. That's a real milestone."

How did they get there?

"We always move pragmatically: we ship a version, even an imperfect one, to test usage. Then we improve it. We'd rather put something in a doctor's hands, even imperfect, than chase perfection. But careful: in healthcare, the 'imperfect' still has to be very solid. A doctor who sees something shaky instantly loses trust."

The arrival of LLMs: opportunity and danger

To standardize and enrich medical data, Klineo also uses advanced AI models.

"LLMs are very powerful, especially for understanding unstructured medical text. They embed enormous medical expertise. That lets you move fast without having a doctor behind every transformation."

But he raises a big warning:

"You should never let a model make a decision when the data isn't there. Never. You can quickly end up with an LLM hallucinating an answer because it thinks it 'should' be there. So the challenge is precisely detecting: does the data exist or not? And if it doesn't exist, who should decide? The human or the AI?"

"The thing that challenged me most was hiring my first devs"

He laughs when he thinks back to that period:

"Honestly, it was hard. I'm not a web developer. I code, I love it, but I come from data. Front-end, back-end, it was all new to me. And I had to hire the first people on the team."

He recounts spending weeks sourcing devs everywhere:

"Someone once told me: a CTO who recruits spends 70% of their time on it. I thought it was a joke. It wasn't a joke."

Why was it so hard?

"First, the dev ecosystem is super varied. Backgrounds have nothing to do with each other. And a great CV doesn't mean a great dev. You have to screen. And screening means talking to everyone."

What he learned:

"Today, I have an eye for it. I've understood which companies have a real tech culture. I've understood which signals show that a person will grow fast. And I have a rock-solid methodology: a question framework, a timing, clear stages."

A recent addition that has helped him a lot:

"Reference calls. I never used to do them. Now they help me decide when I'm hesitating. On my last hire, they let me make the call with more confidence."

A mistake he'd handle differently

"I think we lost time on data analytics. From the start, we were convinced we needed a data-driven strategy to iterate on our product. But we started with hand-built dashboards in Python. It worked, but it didn't scale because there was a barrier to entry (= the code) for contributing, so everything had to go through tech. Then we set up Metabase, which let everyone (product, business, ops, …) put data at the heart of their decisions."

He smiles:

"It's not even that doing it right would have taken more time. It's just that we didn't ask the question. We moved too fast. We knew it would work, so we did it. And in the end, with another technology everything was immediately much easier and instant."

Lesson:

"Always look at the new tools before you start, and talk to others who've been through it. Always."

Staying current: a vital reflex

"Before, keeping up was interesting. Today, it's vital. Everything moves too fast. Every week, a new model comes out. Gemini 3, Llama, OpenAI… You have to stay up to date, or you get left behind in a few weeks."

His favorite resources:

• Hacker News (Y Combinator)

"It's very technical, but extremely rich. I read it religiously every week."

• Podcasts, often while working out:

"I listen to Génération Do It Yourself, Data 101… and actually I even did a Data 101 episode."

• Slack communities

"Modern Data Network, TechRocks… These are gold mines. You have people asking exactly the questions you're asking yourself, and you can search the Slack to find super-specific discussions."

• Conferences

"It used to be AI For Health. Today it's called Adopt AI. It's still great."

Juniors, seniors: two very different challenges

I ask him what makes the difference between a good junior developer and a good senior developer today.

His answer is crisp.

For juniors:

"The danger is leaning on code assistants. It gives you the impression you're moving fast, but if you don't understand what's being generated, you don't learn anything. You have to question everything, re-read everything, keep a strong critical sense."

For seniors:

"The opposite risk: resting on what you know. Telling yourself 'I tried this tech once, 50% of it was bad, so I'm not going back to it.' No. You have to dig in, get up to speed, understand how to use them well. LLMs won't be optional. There will be a real gap between those who adopt them well and everyone else."

What he enjoys most: solving, sharing, understanding

He also talks about his experience teaching at Le Wagon:

"We were building deep learning programs for companies. It was great. Explaining concepts is a good test: if you can't explain them, you haven't really understood them."

And his role as a business angel:

"I love entrepreneurship. I invest in projects whose main theme is healthcare, AI, or impact. Not with big amounts, but enough to help. Often, I make intros, I advise on hiring. And I love talking with founders, following their quarterly newsletters. It keeps you close to innovation."

His values: perseverance, trust, high standards, impact

During a recent offsite, the Klineo team formalized its values. He talks about them very naturally.

"Perseverance is essential. Our users are very busy doctors. Getting their feedback can take weeks. Clinical data is hard. Nothing is simple."

"Trust is essential. Within the team, there's no politics, no competition. And on the user side, we're dealing with hyper-sensitive data. The product has to be impeccable."

"High standards, yes. Even if we ship imperfect versions, the bar stays high. An oncologist won't forgive a mediocre experience."

"And impact… that's our reason for being. With every new feature, we ask ourselves: does it serve 1 doctor or 1,000? Does it really help a patient? If the answer isn't clear, we don't do it."

The team, management… and the joy

He drops a sentence that pretty well sums up his vision:

"We've had a single departure in three years. That makes me really proud."

Why?

"Because in a startup, the team is everything. We spend so much time together. And when you feel that people are growing, that they're happy to be there, that they help each other out… it's the best feeling."

What's next?

"Continuing to make doctors' work as easy as possible so that clinical trials can benefit as many patients as possible. Concretely, that means tackling all the data topics we want to crack, but also the product roadmap, which is huge, and finally continuing to expand internationally."

He thinks for a second:

"But above all, I want us to keep going without losing what we've built in the team. That's the heart of it."

Conclusion: a conversation that goes beyond tech

What stands out in this conversation isn't just Nicolas's technical expertise — even though it's enormous.

It's his ability to navigate between very different worlds: AI, regulation, medical data, product, hiring, the human side.

He talks about search engine optimization and, a few sentences later, about trust, impact, overworked oncologists, patients with no time to lose.

His vision is both demanding and deeply human:

"We move forward in small steps, but every step has to have a real impact."

Klineo is growing fast, expanding into Europe, adopting new AI methods… but what guides Nicolas remains surprisingly simple: learn, share, understand, and help.

A CTO who speaks without pretense, without jargon, without ego.

A CTO grounded in reality.

And above all, a CTO who never forgets why he does all of this: the patients, the doctors, the teams.