Hello!
I'm Ben. I build application-specific integrated circuits and the software to control them.
Get in touch bdatsko@umich.edu
I'm a Computer Engineering PhD student at the University of Michigan working with Professor Michael Flynn in the Flynn Research Group. I design chips capable of solving NP-hard optimization problems faster and more efficiently than Von Neumann architectures by exploiting analog dynamics. I recently completed my Bachelor of Science in Computer Science and Engineering from the University of Michigan.
When I'm not in the lab, I enjoy rock climbing, baking chocolate chip cookies, and tackling bucket list adventures with friends and family.
- C/C++
- Python
- JavaScript/TypeScript
- Verilog/SystemVerilog
- MATLAB
- Assembly (RISC-V, ARM)
- Cadence Virtuoso
- Mixed-Signal Design
- FPGA (Vivado)
- PCB Design
- Embedded Systems
- React/Node.js
- Flask/Django
- MongoDB/Firebase
- FreeRTOS
Three-chip DARPA/NASA platform solving optimization problems 1000x faster than CPUs. Complete web-to-silicon stack for physics-inspired computing.
Physics-inspired SAT solver using 757x less energy than digital SOTA, microsecond solutions, no preprocessing needed. Scales to 50 variables with 100% success rate. Core tech for verification/AI. ISSCC 2025, Nature Electronics.
Error correction decoder using 6-body oscillator coupling. 11x more efficient than digital decoders, beats belief propagation at low SNR. Enables 5G/satellite comms. ISSCC 2024, ESSERC 2025.
Next-gen SAT solver scaling to 200 variables. Dual-tile architecture with integrated RISC-V core. Target: 5000x efficiency vs software.
64-element 28GHz digital beamformer chiplet. 30x power reduction via distributed processing. Streaming-AIB: 1.4pJ/bit, BER 3E-12.
Race strategy platform that helped win 2024 American Solar Challenge. Real-time telemetry, route optimization, weather modeling.
Resume optimizer using NLP to match job descriptions. Automatic tailoring, keyword extraction, Three.js visualizations.
Analyzes Michigan CS/EE courses. Scrapes workload data, grade distributions, enrollment trends. 5000+ course dataset.
Parallel ASIC test framework. 20MHz scan chains, 32KB transfers via Teensy. Cuts validation time 10x.
Power supply controller achieving <10μVrms noise floor. Critical for analog compute research. FreeRTOS-based.
Interactive pathfinding visualizer. Dijkstra, A*, BFS, DFS in pure JavaScript. Used in Michigan algorithms courses.
High-performance typing test. Real-time WPM/accuracy analytics, 1000-word corpus. Built with TypeScript.
Maps NYT articles to Reddit sentiment using Azure AI. Analyzes public perception gaps between journalism and social media.
Official Mojang maps. 250K downloads, 35M YouTube views. Complex command systems, custom AI behaviors.
Swim training app with official start sounds. Syncs visualization to stroke rate for mental rehearsal.