CV
Machine learning researcher focusing on agents for science and engineering.
Technical Skills
- Machine Learning
- PyTorch, JAX, Slime, harness engineering, flow matching, Weights & Biases
- Scientific Computing
- NumPy, SciPy, distributed training (DDP), SLURM
- Chip Design
- FPGA, Vivado, SystemVerilog, SPICE, agents for chip design
Education
Work & Research Experience
Research Engineer
Mar. 2026 - Present- Built domain-specific datasets for semiconductor tapeout and post-training (SFT and RLVR) of GLM 5.2.
- Shipped an agentic analog mixed-signal design verification toolchain used by Broadcom, Micron, and other tier-one semiconductor companies.
- Implemented model routing, programmatic tool calling, and caching optimizations that reduced token use by 40%.
Computational Physics Fellow
Jun. 2025 - Mar. 2026- Developed codes for modeling relativistic electron scattering in dilute gases for nuclear-weapons physics applications.
- Built flow-based models to super-resolve observables in particle-in-cell simulations of fusion reactors, jointly with the Jeong Lab at UCSB.
Undergraduate Research Assistant
Oct. 2022 - PresentMachine Learning Researcher
Jun. 2024 - May 2025- Built flow-based generative models for representation learning of early-universe dark-matter density fields.
- Built flow-based models to super-resolve observables in particle-in-cell simulations of fusion reactors, jointly with Los Alamos National Laboratory.
Physics-Based Computation Researcher
Oct. 2022 - May 2024- Built the first hardware implementation of a higher-order Ising machine, achieving state-of-the-art performance on XOR-satisfiability problems.
- Implemented graph-clustering algorithms to optimize device layouts and enable efficient hardware resource sharing, increasing device capacity by 20%.
Research Intern
Jun. 2024 - Dec. 2024- Built graph-neural-network surrogate models for accelerating LIGO interferometer simulations.
- Achieved up to 800x speedups over traditional numerical methods with minimal loss of fidelity.
- Compared frequency- and spatial-domain laser-cavity simulations, improving the simulation of thermal aberrations.
Flight Software Intern
Jun. 2023 - Sep. 2023- Built data-analysis tooling for disk-space allocation on Astranis satellites.
- Wrote a Global Navigation Satellite System receiver interface for satellite localization during orbit raise.
Firmware Engineering Intern
Jun. 2022 - Sep. 2022- Designed, tested, and documented a lightweight software framework for verification of TenaFe's SSD controller.
- Wrote core modules for simulating error correction, buffers, flash memory, and read, write, and erase operations.
Selected Projects
Mini-GPT
May 2026 - PresentImplemented a GPT-2-style language model trained on TinyStories, including a BPE tokenizer, RoPE, multi-head attention, and mixture-of-experts layers.
Fermiacc
May 2026 - PresentBuilding AI agents for autonomous theory generation and validation in high-energy physics, in collaboration with UCSB physicists.
Publications
- Kannan, S., Qiu, T., Cuesta-Lazaro, C., and Jeong, H. (2025). Lambda-CFM: Scale-Aware Representation Learning for Cosmology with Flow Matching. Machine Learning: Science and Technology, in press.
- Kannan, S., Goodarzi, P., Papalexakis, E., and Richardson, J. (2025). Graph Neural Networks for Interferometer Simulation. NeurIPS 2025 Workshop on AI for Science.
- Nikhar, S.*, Kannan, S.*, Aadit, N. A.*, Chowdhury, S., and Camsari, K. Y. (2024). All-to-all reconfigurability with sparse and higher-order Ising machines. Nature Communications, 15, 8977. *Equal contribution.
- López-Paradís, G., Hair, I. M., Kannan, S., et al. (2024). The Case for Data Centre Hyperloops. Proceedings of the 51st ACM/IEEE International Symposium on Computer Architecture, 230-244.
Teaching Experience
Undergraduate Learning Assistant
Sep. 2023 - Present- Assisted with Intro to Scientific Computing, Electrostatics, Data Structures and Algorithms I & II, and Discrete Mathematics.
- Held office hours and graded assignments.
Course Instructor
- CMPTG CS 5: Statistical Physics and Neural Networks - Taught a ten-week course connecting neural-network theory and statistical physics, including energy-based and diffusion models and neural tangent kernels.
- PHYS CS 5: Non-equilibrium Statistical Mechanics - Taught a ten-week course covering stochastic processes, Brownian motion, kinetic theory, and open quantum systems.
Selected Talks
- Lambda-CFM: Scale-Aware Representation Learning for Cosmology with Conditional Flow Matching, Simons Foundation, Learning the Universe Collaboration invited talk, Dec. 2025.
- Self-Consistent Relativistic Electron Scattering for X-Ray Diagnostics, Kavli Institute for Theoretical Physics, UC Santa Barbara Undergraduate Research Symposium, Sep. 2025.
- Statistical Mechanics of Machine Learning, NSF Institute for Artificial Intelligence and Fundamental Interactions reading group, Mar. 2025.
- Graph Neural Networks for Interferometer Simulations, Kavli Institute for Theoretical Physics, UC Santa Barbara Undergraduate Research Symposium, Sep. 2024.
Awards & Honors
- NSF Graduate Research Fellowship Program Honorable Mention
- Regents' Scholar - Merit scholarship awarded to approximately 2% of incoming UC Santa Barbara undergraduates.
- Semiconductor Research Corporation Research Scholar - One-year undergraduate fellowship sponsored by IBM supporting research on probabilistic computers.