Davis, California, United States
Been watching Sebastian Lague's Game Dev vids. Will start my PhD soon. Currently: Robotics (gantry automation, nanometer precision), Computer Vision, QA/QC, Vacuum and optical systems, Manufacturing processes @ CMU MLOps, Data Pipelines, Modular ML framework design @ Princeton (IRIS-HEP Fellowship) Previous: Di-Higgs analysis, high-throughput parallel processing, big-data pipelines @ CERN (CMS Group, UC Davis) Helium Cycle leak-detection data acquisition -> Linux server -> Grafana/VictoriaMetrics dashboards under Prof. Curro @ UC Davis DLT40: image masking, fringe correction, galaxy/star detection and annotation using Astropy under Prof. Valenti @ UC Davis Tutored: Lower-division: Classical Mechanics, Calculus, Linear Algebra, Differential Equations, MacroEcon Upper-division: Operating Systems, Systems Programming, RISC-V, C, Go, Prolog, Common Lisp Level 1 ITTPC certified by CRLA; managed 40+ tutors for AATC (UC Davis) for one year
Incoming, Fall 2026. Joining the experimental high-energy physics group to pursue collider-based research with a focus on boosted-object reconstruction and jet-substructure signatures for heavy new-particle searches in CMS.
Running full-stack automation on the HGCAL module assembly systems — LabVIEW + NI Vision libraries wired up with Python CV hooks to get down to micron-level placement on silicon hexaboards mounted to sensors, mounted to titanium plates (sandwich-like). Building custom feature-detection + navigation pipelines to drive the gantry with real-time feedback, pushing throughput while preserving alignment tolerance. Layering in QA/QC automation (vision-based inspection, positional verification, grading) and logging every metric into CERN’s prod DB for traceability. Basically stitching together robotics control, vision pipelines, and data bookkeeping into one reproducible flow so we can crank out thousands of modules with nanometer-class consistency.
ML Ops and Infra Engineering: developing an open-source repository of modular, plug-and-play ML pipeline templates for High-Energy Physics, structured around key ML technologies and use-cases, with an MLOps-focused approach to scalable training, inference, hyperparameter optimization, experiment-agnostic design, and production-grade reproducibility.
Working on the Boosted Event Shape Tagger (BEST) project with Prof. Erbacher • Multithreaded the processing pipeline across Particle Class: Splitting, Flattening and Standardizing; reducing processing time by 60–80% in the cmslpc GPU nodes. • Upgraded the collision data procesing pipeline and classifier to be modular with CMSSW v14 from CMSSW v10 for Run-3 Analysis. Understood the CRAB preprocessing code and successfully ran and debugged CRAB-Jobs to obtain collision data encoded in ROOT files. • Currently utilizing Deep Neural Networks to study the composition of Particle Jets from collisions in the Large Hadron Collider. Curating and experimenting with input features like (Transverse Momentum, N-Subjettiness, etc.) using masks and transformations to understand the role of isolating event-level information on network performance. Conducting Hyperparameter Tuning using the tensorboard framework and exploring boosted frame transformations embedded within network layers. • Exploring Ensemble Methods of Classification using Neural Networks to retain and use information in all Class Confidence rates for each Forward Pass (before Softmax processing) • Working with Graduate and PhD students under Prof. Robin Erbacher on research and upgrades for the BEST (Boosted Event Shape Tagger) project.
Overseeing software components of workshops and projects. Developing the official website using React and Next.js. UI/UX using framer-motion and tailwind.css, and some modified components from Aceternity UI. Working on the technical components of workshops, presentations, and outreach events focused on Quantum Information Theory, Quantum Error Correction, Quantum Algorithms, Quantum Cryptography, and Quantum Hardware Implementation.