Breakpoint AI
Founding diffusion research engineer
Breakpoint AI, San Francisco, California, United States, 94199
Direct message the job poster from Breakpoint AI
Computer vision is broken, and we’re going to fix it.
While LLMs have raced ahead, computer vision has been left behind. For basic tasks like object detection, the latest foundation models are outperformed by basic models following the same recipe from 10+ years ago: label a bunch of data and hope that your test set looks like your train set. Unsurprisingly, when applied to problems in the real world these models underperform, requiring end users to spend time and money on domain‑specific, hand‑labeled data in order to get acceptable performance.
The core problem? Getting vision models to work well still requires labeled training data, and getting labeled training data is a pain.
Breakpoint makes it possible to create infinite labeled training data for any use case in only a couple of minutes.
The reason this is now possible is simple — diffusion models. In the last couple of years, the problem of text to image has become nearly solved — we can generate any image we want. At Breakpoint, we are training text to labeled image foundation models that can generate any labeled image you want, and adapt to any use case with a handful of images.
We have a world‑class team with PhDs/postdocs from Berkeley, Stanford and Princeton, and experience at most of the major tech companies (Google Brain, Facebook AI, Amazon, etc.). Our founder, Jamie Murdoch, is a Berkeley PhD with 3k citations, who previously co‑founded and sold a financial AI startup. We are well‑funded, and backed by top‑tier seed‑stage VC firms.
What you’ll do You will help us develop our core image generation technology. This entails
Defining and executing our research agenda, together with the CEO
Devising and running experiments with new diffusion techniques
Staying on top of the latest developments in generative image models, and updating our plans accordingly
Doing whatever else is needed to get our models working better
Help us hire a world‑class team
Qualifications
Experience doing research in deep learning
Strong technical communication skills
Ability to translate abstract AI ideas into working code
Pragmatic and user‑centric — we do not care about writing papers, we are delivering real value for customers
A self‑starter who will do what it takes to get things done as one of the first employees at a venture‑backed startup
Nice to haves
Experience with diffusion models (open source, publications, or just playing around on your own)
Publications at top AI or CV venues (NeurIPS, CVPR, etc.)
Why Breakpoint?
Help shape the future of our company as one of the first employees
Work on cutting edge generative image models
Meaningful equity compensation
Unlimited PTO/sick days
Medical/dental/vision coverage
Well funded from top‑tier investors, with ample runway
Get to work in person from our San Francisco office at least 3 days a week.
We are happy to sponsor visas, and in the past have done sponsored new O‑1As and done H‑1B transfers.
Seniority level Entry level
Employment type Full‑time
Job function Engineering and Information Technology
Industries Software Development
Referrals increase your chances of interviewing at Breakpoint AI by 2x.
Get notified about new Research Engineer jobs in San Francisco, CA.
#J-18808-Ljbffr
Computer vision is broken, and we’re going to fix it.
While LLMs have raced ahead, computer vision has been left behind. For basic tasks like object detection, the latest foundation models are outperformed by basic models following the same recipe from 10+ years ago: label a bunch of data and hope that your test set looks like your train set. Unsurprisingly, when applied to problems in the real world these models underperform, requiring end users to spend time and money on domain‑specific, hand‑labeled data in order to get acceptable performance.
The core problem? Getting vision models to work well still requires labeled training data, and getting labeled training data is a pain.
Breakpoint makes it possible to create infinite labeled training data for any use case in only a couple of minutes.
The reason this is now possible is simple — diffusion models. In the last couple of years, the problem of text to image has become nearly solved — we can generate any image we want. At Breakpoint, we are training text to labeled image foundation models that can generate any labeled image you want, and adapt to any use case with a handful of images.
We have a world‑class team with PhDs/postdocs from Berkeley, Stanford and Princeton, and experience at most of the major tech companies (Google Brain, Facebook AI, Amazon, etc.). Our founder, Jamie Murdoch, is a Berkeley PhD with 3k citations, who previously co‑founded and sold a financial AI startup. We are well‑funded, and backed by top‑tier seed‑stage VC firms.
What you’ll do You will help us develop our core image generation technology. This entails
Defining and executing our research agenda, together with the CEO
Devising and running experiments with new diffusion techniques
Staying on top of the latest developments in generative image models, and updating our plans accordingly
Doing whatever else is needed to get our models working better
Help us hire a world‑class team
Qualifications
Experience doing research in deep learning
Strong technical communication skills
Ability to translate abstract AI ideas into working code
Pragmatic and user‑centric — we do not care about writing papers, we are delivering real value for customers
A self‑starter who will do what it takes to get things done as one of the first employees at a venture‑backed startup
Nice to haves
Experience with diffusion models (open source, publications, or just playing around on your own)
Publications at top AI or CV venues (NeurIPS, CVPR, etc.)
Why Breakpoint?
Help shape the future of our company as one of the first employees
Work on cutting edge generative image models
Meaningful equity compensation
Unlimited PTO/sick days
Medical/dental/vision coverage
Well funded from top‑tier investors, with ample runway
Get to work in person from our San Francisco office at least 3 days a week.
We are happy to sponsor visas, and in the past have done sponsored new O‑1As and done H‑1B transfers.
Seniority level Entry level
Employment type Full‑time
Job function Engineering and Information Technology
Industries Software Development
Referrals increase your chances of interviewing at Breakpoint AI by 2x.
Get notified about new Research Engineer jobs in San Francisco, CA.
#J-18808-Ljbffr