Ethan Fahimi
Astrophysics & Machine Learning
Computer science is no more about computers than astronomy is about telescopes
About
I love numbers; where others see noise, I see a story. I'm a trained researcher in physics, astronomy, and machine learning, with degrees from MIT and The Ohio State University. My passion lies in leveraging patterns within data to drive innovation and deepen our understanding of both the universe and artificial intelligence.
Outside of work, I can likely be found where there are cats, on a run, or speaking about Ohio to anyone who will listen.
Education
The Ohio State University
GPA: 3.99/4.00
Summa Cum Laude
Massachusetts Institute of Technology
GPA: 4.9/5.0
Research Interests
Astrophysics
Using advanced computational tools to conduct analysis at scale and strengthen constraints with some of the most challenging detections in the observable universe. Experience with DECam observations and large-scale astronomical data pipelines.
Machine Learning
Deep learning, optimization, computer vision, NLP, and interpretable AI. Specializing in transformer architectures, gradient boosted trees, and neural networks for scientific applications. Focus on explainability and real-world deployment.
Large-Scale Data
Unsupervised clustering, dimensionality reduction, embeddings (BGE, BERT, SciNCL), and community detection algorithms. Building scalable pipelines for millions of data points using high-performance computing resources.
Experience
Accenture Research
- Conducting applied ML/AI research, developing scalable, interpretable models for real-world impact
- Collaborating across global research teams to advance efficient, data-driven solutions for Accenture’s innovation initiatives
- Leading partnership with MIT Analytics Capstone, bridging academic research and industry applications in machine learning and data science
Massachusetts Institute of Technology
- Applied SHAP explainability for model interpretation to validate hypotheses on drivers of profitability and energy inefficiencies across 450+ warehouses
- Predicted customer profitability for 16M+ items using Gradient Boosted Trees and engineered interpretable features
- Presented at the MIT Analytics Capstone Showcase
The Ohio State University
- Built a scalable pipeline combining SciNCL/BERT embeddings with Leiden/Louvain community detection over 50K+ physics papers to quantify influence and novelty
- Validated the approach on Higgs-boson discovery literature using Python and high performance computing resources
Louisiana State University
- Conducted 12 nights of remote DECam observations on the Blanco 4m telescope of Omega Centauri via NOIRLab, within the MISHAPS collaboration
- Extended time-series photometry and transit-detection pipeline in a maintained Python package; identified candidate exoplanets with methods to scale for 60 DECam fields
- Derived a preliminary upper limit of 0.004 hot Jupiters per star (P = 0.8–3.4 d, 95% CI)
The Ohio State University - GENETIS Collaboration
- Developed genetic algorithms (Python/C++) to evolve antenna gain patterns for neutrino detection via Askaryan effect
- Built a simulation pipeline in Slurm and established upper bounds on detection efficiency in unconstrained geometries
- Contributions resulted in co-authorship of publications in Physical Review D, ICRC2023, ICRC2021