Bangladeshi-led startup Tillion lands Y Combinator spot, outpacing a $300m rival

Tillion, a San Francisco-based startup co-founded by three Bangladeshis, has been selected for Y Combinator's 2026 batch. Y Combinator is one of the world's most competitive startup accelerators.

Reports say only about 1 per cent of applicants get into Y Combinator, whose alumni include Airbnb, Stripe and Reddit.

Tillion's four co-founders are Md Shahrier Islam Arham, Kazi Fardinul Hoque and Abdul Basit Tonmoy of Bangladesh, and Arman Luthra of India.

The company wants AI agents to be able to work on the web as smoothly as humans do.

Much of everyday online work is repetitive: logging into many websites, filling in the same forms, copying data from one page to another. AI agents are being tested on these tasks around the world. But when thousands run at the same time, they often get stuck, held up by slow websites and unstable browser sessions.

Tillion is building browser infrastructure made just for agents. Y Combinator's profile of the company calls it “self-healing”. It can run millions of agent sessions a day with no human watching over them. If one session fails, another picks up the task and carries on.

Customers use the platform for large-scale data collection, automated workflows and round-the-clock monitoring. They can rent many browsers at once and connect them to their own systems with a single line of code.

At the heart of the technology is Fortress, an open-source engine built on Chromium, the engine behind Google Chrome.

The founders say the internet is humanity's greatest creation, yet AI agents are treated as second-class users on it. Since every agent ultimately works for a person, they argue, shutting agents out ends up hurting people.

Browserbase, a US company doing similar work, is valued at about $300 million. Tillion has just four people, yet it has topped ComputeSDK, an open and independent benchmark, several times, ranking ahead of Browserbase.

The company says engineers from more than 170 companies use Fortress. Public PyPI and Docker figures show more than 100,000 downloads. Tillion also says a $1 billion AI lab and a Fortune 500 company use its technology.

Shahrier and Fardinul said nobody expected a team this small to top the benchmark. Running thousands of agents without a single failure seemed even less likely.

Arham has won national championships in debating, science fairs and robotics. He built software that helped thousands of students find jobs, and he built Tillion's browser infrastructure from a hostel room.

Fardinul's A-Level results placed him among the world's top students. He later worked at Amazon Web Services, where he ran hundreds of AI models on a single machine.

Luthra, an olympiad medallist, wrote the remittance algorithm behind a government budget of about $1 billion. A Washington think tank later hired him to work on complex data problems.

Tonmoy won several national olympiad titles and led NDITC, the technology club at Notre Dame College in Dhaka. At Wabash College in the US, he won hackathons, including CalHacks at UC Berkeley. He published AI research as an undergraduate. His open-source model helps robots remember what they see and hear, and one of his models ranks among the world's best on a global benchmark.

Global investment in AI is growing fast, and demand is rising for infrastructure and developer tools as well as finished products. Tillion is part of that shift.

Tonmoy believes AI's next big leap will not be smarter answers but AI that works reliably.

The founders say that if Tillion succeeds, people will win back the hours they lose to repeated logins and form-filling. How far it gets will depend on how reliable and safe the technology proves in real use, and how well it meets customers' needs.

“We envision a world where agents do the repetitive work, and humans can dedicate their time to creativity,” said Arman Luthra.