About Us

We find where models fail to understand humans

Semo is a Tokyo-based applied research lab. We co-research with frontier AI labs to find where models fail to understand humans, and build the data, benchmarks and evaluations that fix it.

Founded by University of Tokyo alumni, we pair research with production-scale data infrastructure, so a finding doesn't stop at a report — it becomes data a model can learn from.

Tokyo

Headquarters

Frontier AI Labs

Co-Research Partners

30+

Languages

UTokyo

Founding Team Alumni

Approach

Find the Gap, Then Close It

  1. 01

    Find

    Co-research with frontier AI labs to locate where models fail to understand humans — what people mean, how they speak, how they act.

  2. 02

    Data

    Targeted data that closes the gaps we find, across speech, vision, video and physical interaction.

  3. 03

    Benchmarks

    Benchmarks that turn each gap into something measurable, so progress is tracked rather than assumed.

  4. 04

    Evaluations

    Held-out evaluations with baselines, showing whether a fix actually worked.

How We Work

Three Principles

Co-research, not outsourcing

We work with research teams as partners, sharing hypotheses and findings rather than only fulfilling a specification.

Evaluation first

Programmes start from a measured gap on an evaluation set. The data mix is designed against that gap.

From finding to data

The team that finds a gap also builds the data to close it, on our own collection, annotation and delivery infrastructure.