Chicago, Illinois, United States
Enhanced hls4ml’s synthesis testing pipeline, migrating to pytest and updating Docker images to support modern HLS tools, ensuring reliability for contributors from CERN, Fermilab, and Northwestern. Saved ~$150k/year for Illinois’ wastewater genomics program by implementing a custom Django-based intake system, improving data flow and statewide reporting. Co-developed and deployed the VISTA genomic dashboard on Azure to visualize COVID-19 variant proportions, aiding health officials in making informed public health decisions. Designed and implemented RAG-based solutions for public health decision support, leveraging Azure AI Services and cloud infrastructure to integrate LLMs with structured data sources.
Developed deep reinforcement learning agents (Proximal Policy Optimization and Deep Q-Network) for automated cyberattack response by learning to identify and remove malicious entities on a Windows 10 host, blocking up to 8 of 10 attack steps while avoiding benign removals in 70% of real-world trials.
Built a hardware accelerator in Chisel to improve data exchange between cores in parallel workloads by implementing a personalized all-to-all protocol on a 2D mesh, achieving 100% correct message delivery and operating within 10–15% of theoretical bandwidth efficiency on mesh sizes up to 6×6, tested on a RISC-V-compatible simulation platform. Presented at the NECST Research Line Fair Event held at NECSTLab in July 2022. Programmed in Scala, benchmarked for different dimensions of the mesh and for different size of the exchanged messages.