Product Pricing Security & data Research Publications Blog About Careers Private Beta Talk to us Download Private Beta
Research People

One researcher. Every paper has an author you can write to.

We are deliberately small. Every published result has a named person who did the work and who will answer a question about the method, rather than a press address that routes to nobody.

The group

Who works on what.

Mohit Bhardwaj

Head of Engineering · Local inference · Applied cryptography · Agent security · On-device retrieval · Measurement & transparency

Works across the Dart Agent research programme: on-device model integration, the semantic index, the identity and post-quantum migration work, the agent permission model, and the measurement studies. Sole author of the ten reports listed below.

Interests
Quantisation under task-specific constraintsHybrid dense/lexical retrievalEvaluation harness designScheduling inference against a render loopHardware key attestationPost-quantum migrationIndirect prompt injectionReproducible disclosure measurement
Publications
Advisors

People who tell us when we’re wrong.

Three external advisors review methodology before publication. They are paid for their time, they do not hold equity, and they are free to disagree with us publicly — two of them have.

Advisor

Prof. Sarah Lindqvist

Systems security, KTH Royal Institute of Technology

Advises on the agent threat model and reviews the injection corpus methodology.

Advisor

Dr. Kenji Watanabe

Applied cryptography, independent

Reviews the identity layer design and the post-quantum migration plan.

Advisor

Rebecca Ansah

Digital rights and policy, independent

Advises on what our transparency commitments should mean in jurisdictions we do not operate in.

Joining

We hire slowly and say what we pay.

Research roles at Dart are engineering roles with a publication expectation. You will ship the thing you measured, which is either the appeal or the objection depending on what you want from a research job.

Publication

You publish what you find

No internal review board can suppress a result for being commercially inconvenient. Three of our ten papers report a negative result about something we hoped would work.

Ownership

Prior work stays yours

We do not assign anything you built before you arrived, and residents keep the rights to what they publish with us.

Scope

Research that ships

Every area lead also owns the corresponding production system. The measurement and the implementation are the same job here.