RaceCTRL

RaceCTRL

Inspiration

Realized that fast data transfer means life or death in an activity like racing.

What it does

Our team was inspired by how Formula 1 combines speed, precision, and data to make decisions in milliseconds. We were all deeply interested in working with high-performance computing (HPCs) and saw racing as the perfect testbed for real-time, compute-intensive simulations. In F1, engineers analyze massive amounts of telemetry but still rely heavily on manual interpretation. We wanted to build a system that could use HPCs and AI together to process, predict, and communicate insights faster than ever before.

How we built it

[Telemetry Feed] ↓ [Surrogate Digital Twin] ← (updates every second) ↓ [Modular HPC Engine – Mojo Kernels] ↓ [Race Simulation Orchestrator (Python)] ↓ [AI Co-Strategist (LLM)] ↓ [Streamlit Dashboard]

Challenges we ran into

Installing packages on the cloud for API setup took 3 hours so we couldn’t fully test Had to figure out how to put animations and 3d models without sacrificing performance on the dashboard Minimized latency between HPC to LLM and back to our dashboard.

Accomplishments that we're proud of

Built a fully functional digital twin system that updates and re-simulates in real time. Created a working HPC-AI integration pipeline that generates live strategic insights. Designed an interface that simplifies complex telemetry into easy-to-understand suggestions. Demonstrated that our platform can extend beyond racing to other high-stakes industries.

What we learned

We learned how to build real-time data pipelines that connect HPC systems with AI inference models, enabling continuous communication between simulation and decision layers. Through experimentation, we discovered the value of event-triggered re-simulation, which significantly reduces computational load while maintaining high accuracy by only reprocessing data when conditions change. Most importantly, we realized the need for human-centered AI systems designed not to overwhelm users with data, but to support professionals in critical, high-pressure environments by presenting clear, context-aware insights that enhance human judgment rather than replace it.

What's next for RaceCTRL

We are planning to use the "telemetry data streaming --> digital twin creation --> HPC --> LLM" workflow in other fields, such as surgery and satellite usage.