SELECTED WORK
A few things I've built. WIP...
Project One
JAX-controller project, using one PID-controller and one Neural Nets-controller. The plants used were a Bathtub plant, a Thermostat plant, and a Cournot market competiotion plant.
Project Two
A from-scratch reimplementation of DeepMind's MuZero, applied to 2048. Three neural networks are trained jointly through backpropagation through time, letting the agent improve entirely through self-play reinforcement learning, without ever being told the rules directly.
Project Three
Analysis and modelling of DNF (Did Not Finish) outcomes in Formula 1 using historical race data. Used to test out models as SVM, Random Forest, LogReg and NNs. Completed as a group project with four others as a part of MECENG278 at UC Berkeley.
Project Four
Simulation of an airport DB, using SQL in "backend" and Python with a CLI as the user interface. Built from a Relation Schema in a DBMS class TDT4145
Project Five
My "home page" for my CV, project overview, etc. Will have additions by time. Also a place to test out interactive/dynamic website features.
Project Six
Computer Vision class final project. Used MONAI as the framework for analyzing 3D medical scans to detect breast cancer. Later used through different DenseNets / ResNets to pick out the best model for submission.