Tyler Crosse

Hi, I'm Tyler 👋

I'm a research engineer building infrastructure for AI safety. I'm a research engineer turned graduate student, currently an MSCS student at Georgia Tech focusing on AI safety. I have 7+ years of industry experience as a Senior Software Engineer and Team Lead, and I'm currently finishing my MSCS at Georgia Tech specializing in machine learning and computing systems. My research focuses on evaluating how models can fail, understanding how neural networks work from the inside out, and building the engineering that makes that research possible at scale.

My Journey

My path to computer science has been anything but conventional. I started with a degree in Biomedical Engineering (VCU), planning to build medical devices. I was quickly drawn to software, shipping production systems, leading engineering teams, and learning how to make complex systems robust and maintainable.

After nearly a decade in industry, breakthroughs like AlphaFold and the rapid scaling of language models crystallized something for me: machine learning wasn't just another tool, it was going to revolutionize science in ways I hadn't imagined. That pushed me to deepen my foundations and focus on alignment: making powerful systems understandable, steerable, and trustworthy. That sent me back to school to pair strong systems skills with rigorous ML.

Research

I'm currently a research fellow in two alignment programs:

I completed ARENA (Alignment Research Engineer Accelerator) in London, where my capstone applied mechanistic interpretability methods (activation patching, steering vectors, path patching) to investigate how moral fine-tuning rewires attention circuits in Gemma-2-2b.

I recently had two papers published. You can find my publications and citation profile on Google Scholar.

Experience

Georgia Institute of Technology

MS Computer Science | 2024 - Present

Specializing in Machine Learning and Computing Systems. Coursework includes Deep Learning, Reinforcement Learning, GPU Hardware and Software, Distributed Computing, and Operating Systems. Built FlashAttention kernels, warp scheduling simulators, and branch divergence analysis tools. Current focus: mechanistic interpretability and ML infrastructure.

Knowable

Lead Software Engineer | 5 years

Led the core engineering team for a legal knowledge platform using search, LLMs, and structured data. Built event-driven microservices on AWS (Lambda/SQS/SNS), OpenSearch, and PostgreSQL; improved CI/CD to cut deployment time by 40%; mentored engineers and balanced architecture with delivery.

Rosetta Stone

Senior Software Engineer | 2 years

Contributed to the language learning platform that's helped millions of people learn new languages. Focused on backend systems and API development, working with international teams across multiple time zones.

Virginia Commonwealth University

BS Biomedical Engineering

Where it all started. Studied the intersection of engineering and medicine, but discovered my true passion lay in the computational side of solving complex problems.

What I'm Into

🧩 Mechanistic Interpretability

Sparse autoencoders, circuits, and tools to open up black boxes so we can understand and steer models.

🎯 AI Safety & Alignment

Building systems that are steerable, reliable, and beneficial by design.

🛡️ Robustness & Evaluation

Empirical evaluation under distribution shift; honesty/helpfulness metrics; reproducibility and measurement.

🔧 Systems for ML

Operating systems, distributed systems, and tooling that make ML research reliable, scalable, and repeatable.

⚡ High-Performance Computing

Squeezing performance from hardware: parallelism, GPU workloads, memory systems.

📚 Learning in Public

Sharing the journey—methods, failures, and breakthroughs—so others can learn alongside me.

⚖️ Economics

Understanding the economic forces that shape our world.

🧬 Computational Biology

Understanding the biological mechanisms that underlie disease and development.

About This Digital Garden

This site is my digital garden—a place where ideas can grow and evolve over time. Unlike a traditional blog, content here isn't organized chronologically but by connections and themes. Some pieces are polished articles, others are rough notes or works in progress.

I believe in learning in public, sharing not just the final results but the messy process of figuring things out. If you see something interesting, wrong, or want to discuss an idea, drop me a line!

Built with Astro. Check out the source code. It would be a little weird to just copy it wholesale.

Sites that inspired this one