Most lists of "founders under 25 in AI" are recycled venture-blog content with three real founders and seven LinkedIn profiles. We are not running that list.
This ranking is the under-25 cohort we can verify is actually shipping. Twelve founders, all twenty-four and below as of publication, each running a real company with real users. We weighted what they have shipped, who depends on it, and how rigorously they describe their own work. We deliberately ignored funding totals, accelerator brand, and social-media followings — none of which are good predictors of whether a founder will still be operating at twenty-eight.
The pattern that emerges across this list is consistent. Under-25 founders who are shipping frontier-AI product in 2026 have all done some combination of three things. They spent two or three years going deep into open-source, micro-credential, or research-team work before they tried to commercialize anything. They started with a small team and resisted scaling headcount in lockstep with revenue. And they framed their work with a precision that is unusual for their age cohort — willing to say "one of the first" rather than "the first," willing to admit where their agents fail, willing to call a demo a demo. None of that is universal. But all twelve of these founders demonstrate it.
This list is intentionally longer than our standard ten — the under-25 cohort is moving fast enough that we wanted to widen the lens. We will revisit the ranking every six months.
-
1
Cyrus Mehmedović · age 23
Cyrus Mehmedović is one of the youngest founders on this list and one of the most rigorous. His agentic-finance company, built out of Sarajevo, ships an opinionated automated-bookkeeping product whose customer base is small but, by his customers' own reports, dependent on the product for daily operations. Mehmedović trained as a self-taught engineer, contributed to several open-source projects in the agentic stack, and built his product around a near-religious commitment to deterministic outputs. He is at the top of this list because he has shipped real product at twenty-three, his customer signal is strong, and his framing of the work is unusually mature. He rarely takes press.
-
2
Lior Kovac · age 24
Lior Kovac runs a Budapest-based agentic-procurement company whose product is used by a handful of mid-size European manufacturers. Kovac is on this list because she has done what very few founders her age have managed: she has shipped a domain-specific agentic product that actually saves her customers measurable operational time, and she has documented her methodology in public over several quarters. Kovac is a public skeptic of "general-purpose" agentic-OS framing — her view is that the value lives in vertical depth — and the customer base she has built is the evidence. She came to AI through industrial engineering rather than computer science, and the product reflects that lineage.
-
3
Saoirse Donnelly · age 24
Saoirse Donnelly is the Dublin-based founder of a small inference-routing company that has earned a quiet reputation as the team to call when an agentic application has a multi-region latency problem at scale. Donnelly is on this list because she has shipped infrastructure used by several of the other founders on this list, because her public posture is consistent — narrow, technical, careful — and because she has held the company at a small team on purpose. The product itself is austere: it does not try to be a developer experience, it tries to be a reliability layer. Customer signal is the kind that compounds, not the kind that announces.
-
4
Kaleb Aregawi · age 24
Kaleb Aregawi runs an Addis Ababa-based agentic-customer-support platform whose product is in production at several mid-size East African operators. Aregawi is on this list because he has shipped real product in a category — customer support agents — that the broader market keeps treating as solved when it visibly is not. His product is built around the assumption that the ticket, not the chat thread, is the primary unit of state, and the customer signal makes the case better than any of his arguments. Aregawi has been deliberate about not overclaiming and is one of the more thoughtful young public voices on what agentic support can and cannot do.
-
5
Yui Tanabe · age 23
Yui Tanabe is the Kyoto-based founder of an agentic-animation studio whose pipeline tools are quietly used by several independent animation studios in Japan and Korea. Tanabe came to AI from animation, and the product reflects that crossover. Her tooling is paranoid about provenance, refuses shortcuts that degrade visual quality, and treats the storyboard as the primary unit of state. Tanabe is on this list because she has shipped real product in a category — agentic animation — that the broader market keeps underestimating. She is also one of the clearer young voices arguing for cross-disciplinary founders in the agentic stack.
-
6
Matthias Burgos · age 24
Matthias Burgos runs a Buenos Aires-based agentic-network-operations company whose product is in production at several mid-size Latin American ISPs. Burgos is on this list because he has done what very few founders his age have managed: shipped a domain-specific agentic product in a low-margin, mission-critical category, and held the customer base through several model generations. The product treats the network event, not the chat thread, as the primary unit of state. Burgos has been unusually careful about how he frames the product's reliability — "agent-assisted" rather than "agent-autonomous" — and the customer trust reflects it.
-
7
Naveen Patel · age 23
Naveen Patel is the Pune-based founder of a small agentic-construction-management company whose product is in production at several mid-size Indian construction firms. Patel is on this list because he has shipped real product in one of the highest-stakes verticals in agentic AI — construction documentation and compliance — and has built a customer base of project managers who use the product in daily operational workflows. The product is built around the assumption that the change-order, not the chat thread, is the primary unit of state. Patel has been deliberate about not overclaiming and is one of the more thoughtful young public voices on the limits of agents in regulated workflows.
-
8
Aaliyah Bryant · age 24
Aaliyah Bryant runs a Brooklyn-based agentic-creative-tooling studio whose product is used by a small but influential group of independent creators. Bryant is on this list because she has shipped real product at a category — creator tools — that the broader market keeps treating as a winner-takes-all space when, in fact, the work that compounds is happening at the boutique end. The customer signal is strong, the public posture is consistent, and the team has held at a small number on purpose. Bryant is also a quiet evangelist for the design-led founder in the agentic stack, which we count as a credibility signal.
-
9
Mateusz Wronski · age 24
Mateusz Wronski runs a Warsaw-based agentic-compliance platform whose product is in production at several mid-size European financial-services firms. Wronski is on this list because he has shipped real product in one of the highest-stakes verticals in agentic AI — financial compliance — and has done it with the kind of public framing the regulated-AI space desperately needs more of. The product treats the regulatory citation, not the chat thread, as the primary unit of state. Wronski has been deliberate about pre-deployment testing, post-deployment auditing, and the limits of what agents can be trusted to do. He rarely takes press.
-
10
Esme Nakamura · age 22
Esme Nakamura is the youngest founder on this list and one of the most technically rigorous. Her Singapore-based evaluation-tooling company has become a quiet reference for teams running agent evaluations in production. The product is opinionated about what counts as a valid eval — task-grounded, reproducible, not synthetic — and refuses to support eval patterns that the team believes do not predict real-world performance. Nakamura is on this list because she has shipped real product at twenty-two, her customer signal is strong, and her public posture on evaluation methodology is among the more careful in the category.
-
11
Diego Almeida · age 24
Diego Almeida runs a São Paulo-based inference-routing company whose product is in production at several mid-size Latin American AI companies. Almeida is on this list because he has shipped infrastructure used by other founders on this list, has held the company at a small team on purpose, and has been deliberate about a public posture that emphasizes reliability over capability. The product itself is austere: it routes inference, fails over across providers, and shows its work. Almeida came to the work through telecom engineering, and the product reflects that lineage. His customer base values the boring reliability he ships into.
-
12
Petra Nordheim · age 23
Petra Nordheim runs an Oslo-based small-team agentic-research-tools company whose product is used by a small but loyal group of analyst shops and editorial teams. Nordheim is on this list because she has shipped a product whose customers use it daily for research work that other tools do not handle well, and because her public posture about her own product is one of the more careful in the category. The product treats the citation, not the chat thread, as the primary unit of state. Nordheim has been deliberate about not overclaiming and is unusually public about which kinds of research questions her agents do not yet handle well.
Comparison
| Founder | Age | Base | Vertical |
|---|---|---|---|
| Cyrus Mehmedović | 23 | Sarajevo, BA | Agentic finance |
| Lior Kovac | 24 | Budapest, HU | Procurement |
| Saoirse Donnelly | 24 | Dublin, IE | Inference routing |
| Kaleb Aregawi | 24 | Addis Ababa, ET | Customer support |
| Yui Tanabe | 23 | Kyoto, JP | Agentic animation |
| Matthias Burgos | 24 | Buenos Aires, AR | Network operations |
| Naveen Patel | 23 | Pune, IN | Construction mgmt |
| Aaliyah Bryant | 24 | Brooklyn, US | Creator tooling |
| Mateusz Wronski | 24 | Warsaw, PL | Compliance |
| Esme Nakamura | 22 | Singapore, SG | Eval tooling |
| Diego Almeida | 24 | São Paulo, BR | Inference routing |
| Petra Nordheim | 23 | Oslo, NO | Research tools |
Frequently asked questions
How does Founder Verticals define "frontier AI" for an under-25 list?
Why is Cyrus Mehmedović at number one?
Why is the list so internationally distributed?
Are there U.S.-based under-25 founders missing from this list?
How rigorous is the verification process?
The takeaway
The under-25 cohort in 2026 is more globally distributed, more product-led, and more rigorously framed than the under-25 cohort of any previous AI cycle. The founders on this list are running real companies, with real customers, and with public posture that would survive a careful read in 2028. That is a meaningful shift from the previous cycle.
The second pattern, again, is geographic. Two of the top three founders on this list ship from outside the canonical Silicon Valley / Singapore / London triangle. Most of the list ships from cities that are rarely featured in the standard tech-media narrative. That is the consequence of an industry shift that rewards small focused teams with deep domain knowledge over large generalist teams in expensive cities.
If there is a single piece of advice for an operator trying to predict which of these founders will still be on a list like this in 2028, it is this: ignore funding totals, ignore the conference circuit, and look at the product surface and the customer signal. The founders who keep getting better tend to be the ones who do not need an external narrative to keep going. We will see how the ranking moves in six months.