Manifold AI Labs
Pushing the frontier of applied artificial intelligence.
ML applications spanning population health to genomics
Healthcare is seeing rapidly-growing and widely-varying uses of ML. Our work has been in ML applications spanning drug discovery to population health. Some recent work includes:
- Biopharma collaboration to develop deep learning based pharmacokinetic/pharmacodynamic models to solve an inverse problem in clinical pharmacology.
- Biotech collaboration to identify the associative effect of single-nucleotide polymorphisms using genomic, transcriptomic, and proteomic data, accelerating time to process 4 million variants from 42 hours to 2 minutes.
- Med device collaboration to personalize therapy at ~20 million patient scale, including causal inference for disease understanding, biomarker identification, and personalized intervention design & deployment.
Virtual assistants using modern natural language models
The NLP research community has produced multiple breakthroughs since 2017 that can be utilized in varied applications. Our work has centered around virtual assistants and intelligent search, which makes it easier to tap into the ~95% of the world's knowledge that's not indexed by Google. Some recent examples include:
- Med device collaboration on patient/provider engagement to resolve questions 10x faster, better, and more cost-effectively.
- Biopharma collaboration on commercial operations to infer and explore provider professional graphs more easily.
- Federal health agency collaboration to simplify access to agency-wide knowledge stores.
Object recognition and data extraction
With rapid advances in data-centric AI, computer vision can now be used to understand digital images with a fraction of real-world data. Our work has focused on synthetic data generation and 3D object recognition methods. Some examples include:
- Ambient computing and pose estimation to enable new human-computer interaction models.
- Image classification and object localization in 3D to streamline inventory operations.
Collaborate With Us
Explore ways of collaborating with us to solve your most complex challenges in AI and data science.
Strategic advice during a period of high-stakes choices on an AI initiative or investment
In some cases, we are asked to collaborate on strategic advice during a period of high-stakes choices on an AI initiative or investment. The context varies from an incoming leader building out a new function to investors evaluating a company. This typically involves bursts of engagement by a senior technical leader as needed for the specific context.
AI prototype within three months by a small team
The most common way for us to start collaborating is on an AI prototype or proof-of-concept within three months by a small team. We are typically working towards a prototype which validates a hypothesis about a high-value application of AI which wasn't feasible five years ago and can be easily demonstrated to other stakeholders.
Multiple projects in a research program lasting one or more years
In some cases, we establish longer-term strategic partnerships involving multiple projects in a research program lasting one or more years.
Explore reflections and talks on our pragmatic approach to applying AI to complex problems in the real world.Read More