At WaveshaperAI, I extended a reusable C++ inference engine into a cloud audio-processing workflow using REST APIs, Docker, relational databases, and AWS Lambda. I also designed model-training and evaluation workflows on Google Cloud Vertex AI.
I fine-tuned NVIDIA Parakeet and Whisper for band-limited HF/radio speech, developing domain-specific augmentation and reproducible WER/CER evaluation pipelines. I partnered with clients and product stakeholders to shape experiments and technical roadmaps.
At TONZ, I translated neural-audio research into shipped VST products and implemented the initial pure-C++ inference engine for Intel and ARM across Windows, Linux, and macOS. The work enabled real-time raw-audio processing at CD quality and higher sample rates.
In independent projects, I built DAFxAtlas, a research landscape with hybrid lexical–semantic search and provenance-aware verification, and publicly deployed a Python/Streamlit learning application that uses an LLM to generate quizzes from the official citizenship guide. My Ph.D. research at the University at Buffalo, SUNY, focused on temporal data exchange and repair.

