Jon Bisila

Principal Investigator & Lead Developer

Leading a team to build a benchmark that evaluates LLM capabilities for C-to-Rust code translation, with a focus on memory safety. As legacy codebases migrate to memory-safe languages, understanding where automated translation works (and where it fails) becomes critical for both security and efficiency, and for assessing the feasibility of AI-assisted translation and modernization of legacy systems. The framework assesses state-of-the-art models and techniques on realistic translation tasks, establishing baselines for what's currently possible and identifying key failure modes.

Lead Research Engineer

Built and deployed an end-to-end NLP pipeline using specialized LLMs for domain-agnostic entity extraction and translation across multiple languages. The system augments a manual analysis process, enabling web-scale data processing with dynamic scaling based on load. Conducted initial research that demonstrated off-the-shelf solutions were insufficient, then designed and validated a custom LLM-based approach that met both performance and throughput requirements for production deployment.