Emmanuel Dumont

About Emmanuel Dumont

Overview: Well-published quantitative researcher, inventor on multiple patents, and entrepreneur with 10 years of experience including conducting and managing research in fast-paced environments, and using machine-learning approaches for large and complex datasets across biophysics, optics, and genomics. Skills- Infrastructure: Google Cloud Platform (Compute Engine, BigQuery, Cloud Storage, Dataprep, AI Platform, Could Functions, Cloud Schedulers, App Engine), Slurm, Elasticluster- Packages/Environments: TensorFlow, sklearn, Git, Jupyter notebooks, Docker, various bioinformatics packages- Languages: Python, R, Shell, SQL, MATLAB.

Experience

1. Associate Professor

   Center For Discovery And Innovation

   May 2018 — Present · Nutley, NJ, US

   Designed and implemented prompt engineering and fine-tuning on large-Language models (LLM) to create a conversational AI doctor able to conduct differential diagnoses and evaluate the emergency severity index of conditions- Reduced the complexity of a quadratic algorithm down to O(n.logn), resulting in 10x more efficient computation of an epigenetic imbalance between the alleles of a genome. Used Shell, Python, and SQL on Google Cloud Platform (Compute Engine, BigQuery, Cloud Storage). Ensured portability by creating Docker images. See publication below- Used deep learning models (Perceptron, convolutional neural networks, recurrent neural networks, Transformer, Graph convolutional neural network) and reinforcement learning (Hidden Markov Model) to analyze an epigenetic imbalance between the alleles among sequencing reads deprived of allelic information, resulting in 10x more surveyed sequencing reads (Manuscript in preparation)- Applied clustering (e.g, HDBSCAN) & machine-learning techniques (e.g, logistic regression, support vector machines, regression trees) on high-dimensionality biological datasets to identify biomarkers for scientific research in R. See Github example below- Analyzed dozens of years of time series of electronic health record data (both structured and unstructured) to decipher patterns in disease susceptibility.

Education

• MINES ParisTech

  Master of Science, Mathematics and Physics

  2002 — 2007

• Columbia University in the City of New York

  Doctor of Philosophy (PhD), Biophysics, GPA 3.8/4.0

  2009 — 2013