Postdoc, Multimodal Machine Learning
New York, NY /
Machine Learning /
Fellowship or Postdoc
At Deliberate, we are building technology that brings more objectivity and precision to the diagnosis and monitoring of mental well-being. We leverage the predictive capability of modern inference models, coupled with thoughtfully acquired and expertly annotated datasets, to reduce subjectivity in psychotherapy and psychiatry.
By joining us, you will help with R&D that brings a level of precision to mental health that is more on par with that seen in other medical areas.
We look for technically skilled individuals who are just as thoughtful about the domain to which they apply their knowledge. Scientists and engineers who realize the power of the tools they deal with, seeking to have a positive impact while at the same time working to mitigate biases and unintended consequences.
We value the unique perspectives provided by individuals with diverse backgrounds and are committed to hiring without bias against age, race, gender, disability, religion, sexual orientation, military service, or other protected status. We strongly encourage minorities, women, individuals with disabilities or other protected status to apply.
In this remote-friendly role, the PhD-graduate research scientist will develop machine learning models based on multiple input modalities — primarily audio and video, but also text and others, such as physiological signals. The postdoc will be advised by Jeff Cohn and/or David Lowe, and will work closely with other researchers, as well as software developers and clinicians “in the field”. Notwithstanding, a considerable level of self-sufficiency is expected: the postdoc should be able to pursue all aspects of the research with little oversight, from literature review, to dataset preparation, to development strategy and model implementation, to statistical comparison between methods, to report/paper writing.
- PhD and peer-reviewed publications in ML, Data Science, or related field.
- Experience developing ML models from scratch using PyTorch and/or TensorFlow/Keras, preferably related to regression problems on sequential inputs.
- Experience with the Python Data Science stack (scikit-learn, numpy, etc).
- Software development maturity in Unix-like, GPU-powered environments.
- Prior work on feature extraction from audio and video — e.g. spectrogram, MFCC, pitch, facial features, action units.
- Experience with AWS products.
- Prior work on sentiment analysis from audio/video.
- Prior work with mental health related datasets (e.g. DAIC-WOZ).
- Prior experience with federated learning.
The position is available remotely. The appointment is for 1 year, with a possible extension. Competitive compensation and benefits are offered.
Deliberate is an early stage, venture-backed, technology startup still operating largely in stealth mode. We believe in the power of diverse teams; not in superheroes. Our team contains a mix of clinical, data science, research and business skills and hails from Latin America, Europe and the US. Most of us have current or past affiliations with leading institutions including Stanford GSB, Stanford CS, Harvard Medical School, Carnegie Mellon CS, Oxford University and Google AI.
We provide objective measurement for behavioral and mental health, serving both healthcare providers and clinical trials. Disorder classification and the subjectivity of their assessment are cornerstone problems in behavioral and mental health, which require solutions urgently. Behavioral health problems are highly prevalent and continue to rise, especially during COVID, with at least 1 in 5 people dealing with some kind of mental health problem. Deliberate exists to identify and improve the condition of affected individuals, and reduce the enormous burden that mental health problems place on society.
Our approach to work and compensation
Deliberate provides competitive compensation and excellent benefits. We believe that a hybrid working model is foundational to both team and individual performance and wellbeing. We are remote-friendly with sufficient time spent in the office to build strong relationship with your colleagues.