ML Researcher at Conduit

About the role

Conduit is a frontier AI research company, training on a new modality of data: brain signals. Our first goal is to decode thought into text through a non-invasive headset. We're a team of about ten ML researchers and software engineers, working together in San Francisco.

Conduit's mission is to increase the communication bandwidth between humans and AIs—and to do so soon. We believe AGI is about to be a solved problem, but keeping humans empowered is not.

We have about 30,000 hours (and counting) of time-aligned brain signals and typing data, from about 1% of San Franciscans. Your job is to 1) train useful models given the data, and 2) steer our data collection operation so we get more useful data.

What you'll do

Train models: We're in a data-limited regime. As we scale up to 100x the amount of data we have now, we hope to also get 10-60x data efficiency wins. Research spans encoders, decoders, synthetic data, multimodal training, scaling laws, evals, post-training, and more.

Steer data collection: As we expand data operations, we want to track where our data collection can be improved. You'll be answering questions like: does sensor X or sensor Y lead to better downstream model performance? If we put the sensors on a dummy head that has known physical properties, do we see sensors performing as expected? Answering these questions involves physical experiments, statistical heuristics, and training many models.

Optionally, a bit of everything: We're a team of about ten. If you want to wear many hats – hardware, fundraising, recruiting, product – we welcome your efforts! If you prefer to focus solely on research, we welcome that too.

Your analyses will heavily rely on AI, and you'll have all the AI tokens you could want.

Who you are

Most important qualities:

1. Brilliant and curious
2. Deep and broad knowledge of ML
3. Bonuses: undergrad-level physics and/or neuroscience intuitions

Example current employees include:

• Former lead of an AI alignment fellowship who studied physics and ML.
• PhD in ML architecture research with a B.A. in math, who questions everything from first principles.
• Top-5 college computer science dropout who joined a frontier AI company at age 21 and had previously interned at a top quant finance firm.

Our bar is a bit higher than the bar for researcher hiring at OpenAI/Anthropic/DeepMind, with the exception that we'll consider candidates with exceptional physics intuitions.

Logistics

• Location: San Francisco, fully in office
• Schedule: Full-time
• Benefits: Health, dental, and vision insurance. Free food for lunch, dinner, and whenever you're hungry. Personal trainer available. We pay for your commute.

Application process

Reach out to Naomi Bashkansky at naomi@condu.it with a resume. We'll get back to you, likely within the day.

Our process:

1. A 30-minute call to get to know each other.
2. A 2-hour coding assignment, where you're allowed to use AI.
3. An hour-long chat about ML and the company.
4. An approximately two-day work trial, where you'll do real tasks alongside us.
5. Offer.