nuray@nurayg.me:~$ whoami

hey, I'm Nuray Ganbarova

I am an engineer building across software and AI/ML, from mobile and web apps to AI agents and computer vision. I have a rather investigative nature when it comes to learning: I like knowing the ins and outs of concepts and the answers to seemingly unnecessary "what if" questions. Once my investigation concludes 🕵🏻‍♀️, building is a breeze.

study: B.S. Computational Sciences (Concentrating in Artificial Intelligence), Minerva University [Graduating May 2027]B.A. Arts & Humanities (Concentrating in Arts & Literature), Minerva University [Graduating May 2027]
seeking: 2027 winter/spring internship, intriguing collaborations, full time positions
based: San Francisco
contact: nuray@uni.minerva.edu
links: resume.pdf github linkedin

# night visits to the 8 priestesses of culture

nuray@nurayg.me:~$ cat interests.txt

I like AI best when it helps professionals in other fields do their work better. That's why I am interested in entrepreneurship: building a startup means figuring out exactly where AI belongs in someone's work.

Outside of tech, I read and write poems, enjoy seeing sculptures from all possible angles, and love discussing art markets.

nuray@nurayg.me:~$ cat experience.md

Google

Software Engineer II (L3) Incoming

Core, Bay Area

Software Engineer Intern May to Aug 2026

Core, Irvine

Built a self-optimizing lifecycle for Gems, custom data analysis agents, that writes their instructions from data assets and keeps refining them from feedback.

stack: TypeScript, Angular, C++, Python

Built a backend generator agent that writes tailored system instructions from data assets (tables, dashboards), turning Gem setup into one click, about 15x faster than writing instructions by hand.
Added a user-triggered mid-chat instruction refresh that analyzes the full chat history and surfaces proposed changes in a side-by-side diff viewer for approval.
Architected a feedback loop where a background agent proposes updates from user feedback and corrections as the chat progresses, scoped to the most recent messages for about 10x lower cost per update.
Wrote it in 3,000+ lines of TypeScript and C++.

photos

Minerva University

Technical Interview Prep Instructor, Sep 2026 to present

Running mock technical interviews and giving actionable feedback to students across 4 cohorts.

Conducting mock DSA, OOP, and System Design interviews in different languages (e.g., Python, C++, Java, etc.).
Giving feedback on problem solving, communication, and code quality.
Guiding students on interview preparation strategy.

Earlier

Robotics research at Universidad Abierta Interamericana in Buenos Aires, building 3D printed robots with IoT sensors and AI chatbots. Software engineering at UNIPORT in Seoul, where a new React flow cut the Alien Residency Card application from 30 days to 5 minutes. Mobile development at Andgreen in Taipei, building an iOS app in React Native before launch. AI in education research at Minerva University.

nuray@nurayg.me:~$ ls Projects/

# the hackathon venue, rebuilt as a room to walk through

MindSpace

Worlds in Action Hack

A walkable 3D mind palace made from memories and notes. Gemini selects the memories that matter most, World Labs builds a Gaussian-splat room around them to explore in WebXR, and PICO links every object back to the note it came from.

stack: Python, Gemini, World Labs, WebXR

links: devpost github

photos

# San Francisco, earthquake and evacuation simulated

SimFrancisco

NVIDIA Spark Hacks

A GPU-accelerated earthquake evacuation simulator for San Francisco, built with NVIDIA RAPIDS. It models risk, debris, and how people move across the city using USGS data, then runs real-time evacuation with many agents at once on GPUs.

stack: Python, RAPIDS (cuDF, cuSpatial, CuPy), GPU Computing

links: github

photos

Failproof Travel

MongoDB Hackathon

An AI travel agent with vector memory and x402 payments. It plans trips over many turns by pulling up flights, hotels, and web evidence in under a second, and skips re-embedding travel data that hasn’t changed.

stack: MongoDB Atlas Vector Search, Voyage AI, x402

links: github

nuray@nurayg.me:~$ cat Research/*.md

Capstone: Recognizing Junji Ito's Style with Deep Learning

with Prof. Patrick Watson, Sep 2026 to May 2027

A deep learning study of artistic style in manga, focused on Junji Ito. Pages from Ito's books and the Manga109 dataset go through Inception V3 with frozen ImageNet weights, the output of each block is averaged into a style signature for the page, and a small neural network classifies pages by artist, Ito vs. other, and genre, following Laubrock and Dubray (2018). To separate his drawing from the horror genre he is known for, I test the model on unseen books, on pages with the text painted out, block by block across the network, and on Cat Diary, his only comedy, after training on his horror alone.

stack: Python, PyTorch, Inception V3

ML Squat Analysis

with Prof. Alex Diamond, Jan to Apr 2026

A computer vision pipeline that reviews squat form from a short video. MediaPipe tracks 33 body points in every frame, and 12 biomechanical metrics, like knee and hip angles and torso lean, drive rep counting and phase labels. Each rep is then classified as correct or as one of five common mistakes, scored, and turned into coaching tips by an LLM. Training a Random Forest on a labeled dataset raised classification accuracy from 39% with rule based thresholds to 88%.

stack: Python, MediaPipe, OpenCV, scikit-learn, Anthropic API

links: writeup

nuray@nurayg.me:~$ ls ~/Elsewhere

I have lived all around the world, through Minerva, where every city gave me new questions, made me comfortable with discomfort, and left me more curious and hungry for growth.

san-francisco

buenos-aires

seoul

taipei

baku

# type help, or keep scrolling

nuray@nurayg.me:~$