Tyler Mestery

• ▹Working at the intersection of AI and Cybersecurity on the Secure CTO & HyperScale AI Teams.
• ▹Developing a machine learning pipeline for generating mitigations against Web Application Firewall (WAF) threats.
• ▹Architecting multi-agent AI systems using DeepAgents and RAG-based LLM pipelines, deployed on Kubernetes and AWS infrastructure.

• ▹Provided technical support for faculty, troubleshooted hardware, software, and network issues in labs and offices.
• ▹Managed IT inventory, system setup, imaging, deployments, and user account/software maintenance tasks.

• ▹Built an end-to-end automated code transformation pipeline that translates PyTorch ML workloads into JAX, coordinating staged validation and generation using Hugging Face LLMs.
• ▹Evaluated instruction-tuned LLMs (Qwen, CodeLlama, Mistral) and validated correctness of generated JAX code.

• ▹Built CodeClip, a benchmark evaluating LLM performance in multi-turn, preference-driven code translation.
• ▹Developed automation scripts and workflows using Docker, AI APIs, and data pipelines for scalable experiments.
• ▹Deployed on Google Cloud to run Piston for automated compilation and unit testing.

TypeScript · React · Express.js · MySQL · Deepgram · Docker · Git

• ▹Engineered a real-time lecture pipeline using Deepgram (English) and Gladia (multilingual) over WebSocket, with in-session AI tools including live Q&A, a jargon glossary co-pilot, and slide-deck sync.
• ▹Integrated Google Gemini to generate structured summaries, flashcards, and graded practice exams from lecture recordings, PDFs, and YouTube videos.

Go · Docker · Linux · SONiC · Bash · CI/CD · Open Source

• ▹Developed a Go-based automation tool to generate docker-compose.yml files from user-defined topology and interface inputs, simplifying SONiC test deployments using docker-sonic-vs.
• ▹Refactored and extended a Go-based generator using the Cobra CLI framework, improving usability, input validation, and extensibility, while contributing enhanced lab setup workflows through Bash automation.