https://dasokol.github.io
dasokol at umich dot edu
248-767-7289

David Sokol

Experience

  • Rentaflop (2022-present):
  • Founder and CEO
  • Solo-founded my startup rentaflop and built it completely from the ground up
  • Rentaflop crowdsources GPUs across the globe to help 3D Artists render animations faster/cheaper
  • Did everything, including Sales, Marketing, Engineering, SEO, Support, etc.
  • Earns income successfully
  • Boston Consulting Group (2019-2022):
  • AI Software Engineer
  • Helped conduct several studies and make recommendations to highest levels of federal government
  • Built a system using Apache Solr to enable an industrial parts distributor to better search its catalog
  • Created an AI messaging tool to write sentences containing KPIs for a major healthcare company
  • Worked with a small team to build the ETL platform supporting BCG's AI group
  • Data Scientist (2016-2019):
  • Built an entity extraction tool to replace the enterprise solution
  • Worked with a small team to ETL a 70 TB streaming dataset
  • Researched the effects of fine-tuning OpenAI's GPT2 deep learning model
  • Helped model when and where world regions lose stability
  • Wrote various other programs and tools for employer
  • Leidos Software Engineering Internship (August 2015):
  • Built software control screens for Ford's Van Dyke plant using the Niagara Framework

Education

  • University of Michigan - Ann Arbor (2014-2018)
  • Bachelor of Science: Computer Science
  • Bachelor of Science: Economics
  • James B. Angell Scholar (2+ terms of all As)
  • University of Michigan Chess Club
  • OpenLoop Team
  • Built a hyperloop pod with a group of students for a SpaceX competition
  • Installed the microcontroller and sensors that measure distance from the track

Research

  • Artificial Intelligence Research (2017-2019):
  • Research into the impact of adversarial machine learning on insider threat
  • Built a Bayesian spam filter with over 97% accuracy (after 10-fold CV) on the trec 2007 dataset
  • Built a reverse-engineering model using 300000 nytimes articles from the UCI ML repository
  • Done in conjunction with an AI professor at Georgetown University

Skills

  • Tools: NextJS, React, AWS (EC2, S3, Lambda, RDS, ECS Fargate, etc.), Apache Solr, Flask, Stripe, Docker, Kubernetes, git, NVidia GPUs
  • Languages: Python, C/C++, Java, JavaScript, SQL, x86/ARM Assembly, HTML, CSS, Lisp
  • Machine Learning: Transformers, LLMs, GPT, RNNs, CNNs, GANs, SVMs, Naive Bayes, K-Means, LDA, Regression, GMMs, HMMs, Decision Trees, Collaborative Filtering, PyTorch
  • Environments: Linux, Unix, Windows

Previous Education

  • Walled Lake Western HS - Walled Lake, MI (2010-2014)
  • 4.37 GPA
  • Founder and president of debate club
  • Graduated summa cum laude, top ten of class
  • 2350/2400 SAT score (top .1 percentile)
  • Inducted into academic hall of fame

Interests

  • Chess
  • Boxing
  • Chessboxing (yes, it's a thing)
  • Artificial Intelligence, Machine Learning
  • Programming
  • Solving puzzles
  • Piano
  • Violin
  • Running

Independent Projects

  • Selfie Filter
  • Use AI to change how your selfies look
  • Music Visualizer
  • Lights LEDs to music using fast fourier transforms run on an arduino
  • AI Art Generator
  • Generates unique abstract art using PGAN, trained on an Nvidia DGX-1 GPU supercomputer
  • Sudoku Solver
  • Solves the world's most difficult (or any) sudoku puzzle
  • HackMIT 2016 Hackathon
  • Created a decentralized social network using a distributed hash table
  • RSS Feed Reader
  • Parses XML files for real-time weather, sports scores, news
  • ChatterBot
  • Holds conversations with the user (but doesn't quite pass the Turing Test)