Victor Gusev

Reinforcement learning applied to physical systems, from PPO training pipelines in simulation through validated hardware deployment.

Education

University of California, Los Angeles

Expected June 2028

B.S. Computer Science and Engineering

GPA 3.64 / 4.00

Coursework: Algorithms and Complexity, Software Construction, Object-Oriented Design, Linear Algebra

Programming Languages:: Python, C++, C, Java, JavaScript, TypeScript, Assembly
Robotics & Embedded:: ROS, NVIDIA Isaac Lab, ESP32, PID Control, LiDAR, Sensor Integration, Computer Vision
Machine Learning:: PyTorch, Reinforcement Learning, PPO, Domain Randomization, Sim-to-Real
Web & Backend:: React, Node.js, Express, Socket.io, MongoDB, Next.js
Tools & Hardware:: Git, Docker, Linux, SolidWorks, 3D Printing

Software Engineering Intern | Dimaag.ai

June 2026 – September 2026

Developed a hybrid reinforcement-learning and classical local planner for autonomous vehicle trajectory tracking on complex paths
Validated and deployed autonomous navigation policies on physical hardware, achieving reliable tracking at >10 mph with sub-decimeter trajectory accuracy
Implemented PPO algorithm training pipelines and 2D LiDAR sensor raycasting within NVIDIA Isaac Lab
Bridged the sim-to-real gap by validating policies across various simulators, implementing domain randomization to ensure seamless hardware deployment

Robotics Programming Lead | FIRST Robotics

August 2021 – May 2025

Led software development for a 40+ member FIRST Robotics team, building semi-autonomous competition robots
Implemented PID control systems, vision tracking (Limelight), and LiDAR-based localization
Engineered autonomous routines and vision pipelines that propelled the team to the FIRST World Championships, ranking in the top 5% of regional competitors

Proof | software

2026

Architected a full-stack progressive web application (PWA) with MongoDB and React to version-control recipe iterations and manage real-time inventory states
Integrated the Gemini API and web scraping (Cheerio) to autonomously extract, intelligently restructure, and import recipes directly from raw URLs
Engineered a real-time, cross-device timer synchronization system using Socket.io to track concurrent baking stages across multiple platforms

TaskAble | UCLA HOTH

2026

Built an educational web app linking student interfaces to a teacher dashboard through Firebase Firestore, developed with a small team during UCLA HOTH
Integrated the Gemini API to generate task breakdowns that decompose assignments into manageable steps for students
Implemented a reward system and emotion logging to track student engagement alongside task completion