Yusuf Yilmaz

ML, data science, full-stack systems, and research computing.

Welcome

University of Houston - CS + Mathematics

Computer science and mathematics work across research computing, applied machine learning, and full-stack systems.

B.S. candidate at the University of Houston, double majoring in Computer Science and Mathematics with the Data Science option. Project work spans forecasting, regression modeling, reinforcement learning, publishing systems, and Java game systems.

Experience

Research Computing Support Technician

University of Houston, Sen Receptorology Lab

Dec 2023 - Sep 2025

Managed Linux/HPC research infrastructure across NAS/NFS-backed lab systems, including user accounts, access controls, environment modules, and administration of RHEL/AlmaLinux workstations.
Resolved kernel, driver, and XFCE configuration issues on legacy NVIDIA hardware and built a Python script for account synchronization.
Deployed, configured, and debugged GPU-accelerated scientific software stacks including CryoSPARC, Relion, ChimeraX, Phenix, Amber, NMRPipe, CCP4, NVIDIA drivers, and CUDA toolkits.
Designed an OpenCV-based computer vision pipeline for automation of grid detection and handling pipelines in cryo-EM sample preparation robots.

STEM Intern, Instructor, and Facilitator

University of Houston, ITECH STEM & BioTECH Camps

Jun 2019 - Jul 2025

Progressed from STEM Intern to Instructor / Facilitator, supporting computing, 3D printing, and biotechnology programs across multiple summer cohorts.
Led 3D printing instruction covering CAD design, slicing workflows, and additive manufacturing fundamentals; printed personalized projects for 100+ students.
Delivered biotechnology camp instruction on molecular docking workflows using AutoDock Vina and OpenBabel while troubleshooting software setups, lab tools, and classroom equipment.
Partnered with faculty and program staff to iterate on curriculum design, improve instructional materials, and align activities with learning objectives.

Technical Skills

Languages

Python

Java

SQL

TypeScript

JavaScript

Bash

ML / AI

PyTorch

scikit-learn

LightGBM

OpenCV

NumPy

pandas

Matplotlib

Gymnasium

Systems / Tools

Linux

RHEL

AlmaLinux

Git

GitHub

NAS/NFS

SSH

CUDA

Environment Modules

Web / Data Infra

Next.js

React

Tailwind CSS

PostgreSQL

Prisma

Vercel

OpenAI API

Projects

Reinforcement learning

Deep RL Lunar LanderMar 2026 - Apr 2026

Developed and compared REINFORCE, entropy-regularized Actor-Critic, DQN, and Actor-Critic-guided DQN agents for LunarLander-v2.

Built deterministic seeding, resumable checkpoints, and seed-sweep tooling for reproducible RL experimentation. Achieved up to 148 best 100-episode average reward with Actor-Critic and 254 with DQN across seed sweeps; strongest peak performance came from Actor-Critic-guided DQN.

Time-series ML

ERCOT Load ForecastingDec 2025

Built a 24-hour-ahead forecasting pipeline using ERCOT load data and weather from 8 zone proxy locations.

LightGBM achieved R2 0.861, RMSE 3816, and MAPE 4.77% on the 2025 holdout year.

Houston Property Value Prediction project visual

Regression modeling

Houston Property Value PredictionFeb 2026

Compared Linear Regression, Random Forest, and LightGBM on Houston housing data.

The best model achieved R2 0.791 with RMSE 149,491 after trimming the top 1% of price and area outliers.

Full-stack AI product

fortune.tipsNov 2025

Built and deployed a full-stack Next.js app integrating OpenAI-generated tarot readings, motion-triggered interactions, and Server Actions.

Server Actions support streamlined frontend-backend workflows.

Commissioned platform

outliers.social

Dec 2025 - Present

Developing a commissioned text-first publishing platform with role-based access, draft editing and publishing flows, tags, stars, and relational schema design.

Implemented core auth, user/profile, permission, and standalone publishing systems, with publication workflows and moderation/reporting still in progress.

Java game systems

Minecraft Mod DevelopmentJun 2021 - Oct 2023

Designed and released Minecraft mods with 366K+ combined downloads across Modrinth and CurseForge.

Built custom item, block, entity, and gameplay systems, including configuration-driven music behavior and event-driven gameplay hooks. Published and maintained multi-version releases across Fabric and Forge mod loaders, iterating on compatibility fixes and gameplay updates.