Research at Adept

In the past decade, progress in AI and its societal and economic impact have been propelled largely by advances in neural networks. Innovations in deep learning from years of industrial and academic R&D underpin a multitude of useful technologies such as translation, protein folding, self-driving, and programming assistants. Impressive increases in scale combined with simple training techniques now endow a single model with remarkable competence in several tasks, a departure from the bespoke deep learning models of the past that were engineered for specific tasks.

Adept is building generally intelligent natural language interfaces to the digital world, empowering people to interact with machines just like they would with a teammate. At the heart of this human-machine partnership lie some of the most exciting research challenges in AI.

We believe that many of the important research problems on the path to building generally intelligent interfaces can only be accessed with real people using our models and improving them with feedback. To act in the digital world at the users’ behest, our models need several skills:

  • They need to process large amount of contextual information,
  • Understand and manipulate multimodal information,
  • Learn from user feedback,
  • Teach themselves how to use software tools with limited human supervision.

At Adept, you will be working with some of the leading researchers in areas such as modeling, systems, reinforcement learning from human feedback, data curation, and user interfaces. As we build useful products out of capabilities that are already within grasp, our fundamental research will unlock the next tier of capabilities that will further empower users. We have the rare privilege of shaping how people will interact with the digital world and we are looking for passionate and brilliant minds to join us on this journey.