Sohum Sukhatankar Resume

Education

Harvard University

2027 — 2028

Harvard University

2024 — 2028

Undergraduate technical coursework:

  • CS 50: Intro to CS
  • CS 61: Systems Programming
  • CS 120: Algorithms
  • CS 1810: Machine Learning
  • AM 226: [Graduate] Computational Neuroscience
  • MIT 18.337: Parallel Computing and Scientific ML
  • Math 21a: Multivariable Calculus
  • Math 21b: Linear Algebra
  • Math 55b: Honors Topology, Real Analysis, and Complex Analysis
  • Math 123: Algebra II - Theory of Rings and Fields
  • Math 155R: Combinatorics
  • Stat 110: Probability
  • Stat 111: Statistical Inference

Professional Experience

Software Engineering Intern at Newline Interactive

June 2025 – August 2025 Dallas, TX

  • Created internal RAG chatbot to assist IT team using OpenAI embeddings, ChromaDB, and semantic search to deliver contextually-accurate responses with under 1s query latency and 90+% accuracy on benchmarks
  • Architected full-stack chatbot features, employing a React frontend and scalable PostgreSQL-based backend for logging 1k+ chat sessions and 10k+ feedback entries
  • Engineered AI-based OCR pipeline to SAP, parsing purchase order PDFs into an SAP-readable payload, transmitting them to SAP via a secure REST endpoint, and reducing manual data entry time by 85%

Software Engineering Intern at Alibre, LLC

May 2025 – June 2025 Remote

  • Engineered end-to-end design pipeline, leveraging Stable Diffusion with custom text-to-image prompts to synthesize images for downstream CAD workflows
  • Wrote code for PNG-to-SVG conversion using Pillow, OpenCV, and NumPy to detect contours, and enforce creation of non-self-intersecting B´ezier paths for successful extrusion
  • Employed simplification algorithm to render input images of up to 1024x1024 resolution in under 3s

Researcher at Madsen lab, Boston Children's Hospital

January 2025 - May 2025 Boston, MA

  • Developed a machine learning pipeline prototype to predict seizure onset localization strength by integrating electrode spatial data and patient clinical metadata
  • Created graph embeddings using node2vec to capture local and global patterns in directed graphs representing networks of Granger causality
  • Trained a PyTorch-based GNN, with performance benchmarked against a random forest baseline

Awards and Honors

Mathematics:

  • AIME: Qualified x2

Other:

  • Scripps National Spelling Bee: Participant (2017, 2018), Champion (2019), Highest written test score (2019)
  • National Academic Quiz Tournaments (Quiz Bowl): Small School Nationals champion (2024), High School Nationals runner-up (2024), Intercollegiate Nationals DII champion (2025)