Informatics @ UW · Data Science track

Yuta Banishky

Data Analyst

My passion is solving real-world problems with data and AI — not just building things that work in theory, but things that actually make people's lives a little easier. Mainly working in Python, SQL, pandas, Power BI, and Tableau.

Download Résumé Email GitHub LinkedIn

16 projects built and shipped
340+ weekly food bank orders analyzed
18% fewer stockouts on high-turnover items
12% less overstock waste in one quarter

Work

Software & Product 13

2026

Dubly

A husky-themed study buddy wired to your UW Canvas — chat to see what's due, ask about any syllabus, and generate quizzes, flashcards, and study plans from your course material. Fixing a SQLite concurrency bug across 10 data stores with per-thread connections and WAL mode cut the error rate under load from 35% to 0%, verified by a TDD stress suite; token-level accounting put each turn at $0.0475 and identified prompt caching as a 74% cost reduction.

Python
FastAPI
SQLite
Next.js
Claude API
Whisper

2024

AI Debate Colosseum

Set a topic and watch two language models argue it out round by round, each responding to the other's last point.

LLMs
Python
Web

2026

AI Poker Room

Watch OpenAI, Anthropic, Gemini, and Grok models play five-card draw against each other — live betting, bluffs, and showdowns, with each model's reasoning streamed in as it decides whether to hold, raise, or fold.

Next.js
TypeScript
Multi-LLM

2024

AI Resume Analyzer

Upload a résumé and get structured, section-by-section feedback back — a Gradio app running on the OpenAI API.

Python
Gradio
OpenAI API

2025

CineMate

A movie night decided for you, staged inside a retro cinema. Filter by genre, mood, runtime, rating, streaming service, and decade, then let it play titles across the big screen — auto-skipping like channel surfing — with a watchlist, star-rated history, and trailer links. Accounts and avatars run on Supabase Auth and Azure, with a one-tap guest mode.

Next.js
TypeScript
Azure SQL
Azure Blob Storage
Supabase Auth
TMDB API

2024

Charades vs. AI

Act out a word to your webcam and the AI tries to guess it — charades against a computer-vision opponent.

Claude API
Computer Vision
Webcam

2025

Yutachess

Real-time multiplayer chess — share a 6-character room code, pick a time control (Bullet/Blitz/Rapid), and play with live rematch support synchronized over WebSockets. The backend is server-authoritative, validating every move and syncing state between players; load-tested to 3,000 concurrent games and 374K+ moves with zero state desynchronization and 1.89 ms p50 latency.

Node.js
Express
Socket.IO
chess.js

2023

AdSnap

Point your phone at any product and about two minutes later you have a finished commercial — script, cinematic shots, slogans, voiceover, and music.

Mobile app
React Native
Azure AI

2022

KitNations

A Next.js e-commerce store for World Cup 2026 national-team kits from all 48 nations, with product browsing, cart and checkout, and worldwide shipping.

Next.js
React
E-commerce

2023

recipe.AI

Photograph the inside of your fridge and Claude Vision reads the ingredients, then suggests dishes with cook times, nutrition, and full recipes.

Mobile app
React Native
Claude Vision

2022

Hand Tennis

Browser tennis you play with your hands — webcam hand-tracking moves the paddle, no controller needed.

Computer Vision
Hand Tracking
JavaScript

2023

Zyde

A full-stack site I designed, built, and tuned for search — live at zyde.org.

Web Design
Full-stack
SEO

2026

Where My Taxes Go

Pick a city and enter an income, and it breaks your federal, state, and local tax into the programmes it actually funded, two levels deep, in plain English — combining all three levels of government into one personalized picture across six cities. Every number is sourced: a build pipeline regenerates the entire data corpus from primary documents on every run and refuses to write anything if its verification pass fails.

Python
Data Pipeline
JavaScript
Open Data

Data & Analysis 3

Bar chart of median Seattle Priority 1 police response time by precinct, with North precinct above the 7-minute standard

Seattle 911 Response Times

Seattle's 911 dispatch data promises officers reach Priority 1 calls in under 7 minutes. I cleaned 43,773 of 50,000 sampled calls — dropping test entries, bad timestamps, and outliers over 120 minutes — and found the North precinct alone missing that standard at a 7.2-minute median, while the rest of the city met it. A phase-2 logistic regression on 10,770 calls then predicted whether a Priority 1 call would clear in time from call type, geography, and time of day, scoring 0.723 ROC AUC on a chronological holdout against a 0.519 baseline.

Python
pandas
scikit-learn
Logistic regression
Open Data

Line chart of average Seattle crash severity by hour of day, peaking in the early morning and evening commute

Seattle Traffic Collision Analysis

I live in Seattle, and I wanted to figure out where the city should actually be spending its safety budget. So I analyzed 221,266 traffic collision records to find out. Instead of just counting crashes, I built a scoring system that weighs how severe each one was, which showed that cyclist crashes along the Burke-Gilman Trail were a bigger problem than raw numbers suggested. I also checked whether missing streetlights were a cause, since that seemed like an easy fix, but the data showed they weren't.

Python
pandas
Matplotlib
Open Data

2025

AI CSV Data Analyzer

Upload a CSV, ask a question in plain English, and get back the chart that answers it. A three-stage pipeline decides what to compute, runs the pandas itself, then explains the result — and only the column names ever leave your machine, so the raw rows stay on-device.

Python
pandas
Gradio
OpenAI GPT-4o-mini

About

Hello! I'm Yuta, a B.S. Informatics student at the University of Washington on the Data Science track, expected to graduate in June 2029.

I'm drawn to the part of analysis that actually changes a decision. Volunteering at the University Food Bank, that meant working through distribution data to find which items kept running out — and cutting stockouts on them by 18%. Analyzing Seattle's traffic collisions, it meant weighting 221,266 records by how severe each crash was instead of just counting them, which surfaced a cyclist hotspot that raw crash counts had buried.

Before college I founded Bellevue Coding, a free weekly Python program for kids ages 8–14 at the Bellevue Library, and Wilburton Soccer, a youth training program I built and ran from scratch. I still build software — mostly tools that put data or AI in front of people — and most of it is live and linked below.

Currently I'm taking coursework in data programming and statistics at UW, and spending my own time getting sharper with SQL.

Tools I reach for

Python
SQL
pandas / Matplotlib
Power BI
Tableau
Excel
Statistical analysis
Next.js / React

Experience

Professional Experience

Jan 2026 – Present

Data Analyst · Seattle University Food Bank

Analyzed 340+ weekly orders in PostgreSQL to surface demand patterns, cutting stockouts 18% and overstock waste 12% in one quarter.
Built a logistic regression model in Python to forecast restocking by item category; deployed it as a REST API handling ~1,000 requests/day for 200+ households weekly.

Python
PostgreSQL
Logistic regression
REST API

Jan 2025 – Jun 2025

Founder & Instructor · Bellevue Coding

Founded and taught a free weekly Python program for ages 8–14 at the Bellevue Library, growing enrollment 133%.
Extended the curriculum to ML and analytics pipelines for ages 13–17.

Python
Machine learning
Analytics

Leadership & Activities

Sep 2025 – Present

Project Officer · UW Data Science Club

Led a 5-person team building a linear regression pipeline on 2,000+ cleaned HUD and Census records, finding an R² of 0.72 between homelessness and income across greater Seattle, visualized in Tableau.

Python
Linear regression
Tableau
HUD / Census data

Mar 2025 – Present

Member · Machine Learning @ UW

Built a real-time sentiment pipeline for Reddit comment streams with Python and PRAW, fine-tuning DistilBERT and deploying via FastAPI on AWS EC2 with a Streamlit dashboard.

Python
PRAW
DistilBERT
FastAPI
AWS EC2
Streamlit

Résumé

Education

University of Washington

B.S. Informatics, Data Science Track

GPA 3.75 · Dean's List · Seattle, WA

Expected June 2029

Data Programming
Statistical Reasoning
Advanced Statistical Methods
Foundational Skills for Data Science

Achievements

Selected for the UW-IT Student Innovation Lab's Spring 2026 Community Showcase for Dubly; featured on the Lab's website and LinkedIn.
Earned the Microsoft Power BI Data Analyst Professional Certificate, 2026.
Founded two community organizations: Bellevue Coding and Wilburton Soccer.

Technical Skills

Languages: Python, SQL, JavaScript (ES6+)
Data & Analytics: pandas, Power BI, Tableau, Excel, DAX, logistic and linear regression, PostgreSQL, SQLite
Tools: FastAPI, Node.js, Next.js, React, AWS, Git, Claude API, REST APIs

The one-page PDF Everything above plus the full experience and project write-ups, formatted for applications.

Download PDF

Get in touch

I'm looking for data analyst internships. If you have one — or just want to talk about a dataset — my inbox is open.

yutabanishky@gmail.com github.com/YutaB123 linkedin.com/in/yuta-banishky