Kay Krachenfels | UC Santa Barbara

About Me

Hi! I am a Master’s student in the Computer Science department at UC Santa Barbara.

I received my Bachelor’s degree in Computer Science and Engineering from UC Davis. I have both technical software internship and research experience, and I also love engineering that gets more hands-on; broadly, I am interested in using AI for medicine, language/computer vision and low-level hardware (e.g. CUDA), and in software applications such as in automotive and avionics.

Past Experience

[Master’s Project - 2026] Agentic AI Search and Feedback Strategies for CUDA Kernel Generation
- Compared effectiveness of different agentic AI search heuristics (e.g. beam search, MCTS) when optimizing CUDA kernels and studied robustness of various feedback heuristics, such as to non-tested shapes or non-target hardware
- Analyzed realistic impact and speedup of optimized kernels when placed in end-to-end tests in ML/LLM models
[Summer 2025] Avionics Software Intern at Blue Origin
- Worked with Python, C++, Linux, and CI/CD pipelines to create tooling for measuring technical performance and resource utilization of applications, in addition to writing other supportive tooling and contributing to Blue Ring flight software.
[Sept 2021 - June 2024] Research assistant in the Janata Lab at UC Davis
- Used PyTorch to build and train a CNN on 200 music-evoked memory passages for prediction of lifetime periods; used other natural language libraries to analyze/visualize sentiments, similarities, and various patterns across memories.
- Used Django and an AWS-hosted SQL database to set up models and pipelines for running and storing analyses.
- Led project and taught other undergraduate students to get set up with coding environments and database.
[Summers 2023 and 2024] Software engineering intern at General Motors
- Wrote an in-house Python data logging library + GUI to send with all 30+ vehicle electrical control units (ECUs) via DoIP and CAN protocol; allowed users to send requests to any module and decode various diagnostic responses.
- Wrote scripts for real time testing (RTT) scenarios for vehicle simulation and maneuvers; created a plotting tool to visualize relevant signals when verifying bench functionality tests.
[June 2022 - June 2023] Biochem data science/research with Lawrence Livermore National Lab (LLNL)
- Wrote Python scripts and used PyTorch to train, tune, and visualize graph-based machine learning models on 2D SMILES data, including SARS-CoV-2 data, for unsupervised/semi-supervised learning and supervised property prediction. Utilized LLNL’s high performance computing (HPC) clusters to deploy SLURM batch jobs.
- Informally presented results comparing results of training based on proportions of unsupervised and supervised training data used.

Publications

A Compact Handheld Sensor Package with Sensor Fusion for Comprehensive and Robust 3D Mapping

Wei P, Fu K, Villacres J, Ke T, Krachenfels K, Stofer CR, Bayati N, Gao Q, Zhang B, Vanacker E, et al.

Sensors 24, no. 8

PDF Capstone project

Other Work

I play and compose music sometimes! You can check out my music youtube channel.

I did the music & audio for an alternative controller video game, Sashimi Slayer; you can download the game from steam as well as learn more about the game and our team!

[August/September 2025]: Sashimi Slayer was accepted into PAX West (Seattle, WA) as part of the PAX Rising showcase, and we will be exhibiting it August 29th through September 1st
[July 2025]: Sashimi Slayer was exhibited at Secret Sauce as part of Open Sauce in San Mateo
[March 2025]: Sashimi Slayer was exhibited at alt.ctrl.GDC as part of the Game Developer’s Conference (GDC) in San Francisco