Erik Gafni

Recruited and led a team of ~7 machine learning research engineers building various machine learning models and software in LLMs (Nouns, largest DAO and finance), computer vision (DeepCell $70M+), generative vision (founders of MySpace), reinforcement learning and time-series (finance), audio/voice (Sanas.ai $50M+), MLOps, Cloud Infra, etc.

I am very hands-on with code.

Helping the founders of MySpace with generative AI

• Applied computer vision, NLP, and audio neural network architectures to different types of high throughput genetic sequencing data

• Wrote R&D and production software infrastructure for MLOps, cloud, and data pre-processing
• Created a early detection cancer-screening test
• Research results led to $9.5M in funding
• I led a collaboration with the University of Missouri which resulted in a Nature publication.

Machine Learning algorithm development for Invitae's cell-free DNA non invasive prenatal screening test which was deployed to millions of patients.

• Machine learning for the early detection of colorectal cancer from multiple analytes in the blood.
• ~30th employee at a start that went on to raise around $1B dollars in VC.

• Developed and trained probabilistic, machine learning, and bioinformatics methods and production software to process petabytes of clinical genomic data into predictions used for clinical genetic testing reports. Invitae was one of the first biotech diagnostic companies to deploy a machine learning algorithm into production.
• ~30th employee at a startup that IPOed a few years later ($6.5B+)
• Lead development of Invitae’s clinical production variant calling pipeline.
• Author of 2 of Invitae's foundational patents.
• Author of the open source workflow manager that Invitae's pipelines are written in, Cosmos. The library is used by various genomics group around the world to do scientific distributed computing.

• Applied Bioinformatics and Machine Learning methods to NGS genomic data and autism data
• Developed generative probabilistic graphical models (PGMs) for clinical trial simulations
• Worked with one of the first groups (headed by Dr Tim Yu from Boston's Children Hospital) to ever do a clinical exome sequencing and interpretation to end a patient's diagnostic odyssey.
• Wrote an open source bioinformatics workflow manager still used by academic and commercial labs around the world.