Albert Bogdan
Statistics & Computer Science @ UIUC
|
hi, my name is albert, and welcome to my page!
- rising junior studying CS+Stats at a top-5 CS university
- interests in full-stack, data, AI, ML
- B2 biliteracy in 🇵🇱 and seal of biliteracy in 🇩🇪
Stack
Languages
TypeScript
JavaScript
Python
Java
C++
C
R
Bash
Swift
TypeScriptJavaScriptPythonJavaC++CRSQL
BashSwift
Frontend & Backend
React
TailwindCSS
Vite
Node.js
FastAPI
Flask
Spring Boot
TanStack
JWT
ReactTailwindCSSViteNode.jsFastAPIFlaskSpring BootTanStackJWT
Databases
PostgreSQL
MySQL
Cloud SQL
PostgreSQLMySQLCloud SQL
AI & Data Science
NumPy
Pandas
Scikit-learn
Jupyter
OpenAI
Gemini
Claude
Matplotlib
OpenCV
Pydantic
NumPyPandasScikit-learnJupyterOpenAIGeminiClaudeMatplotlibOpenCVXGBoost
PydanticFAISS
Infrastructure & Cloud
AWS
Docker
GCP
Linux
GitHub Actions
Git
Cloudflare
Vercel
Jira
Playwright
AWSDockerGCPLinuxGitHub ActionsGitCloudflareVercelJiraPlaywrightBeautifulSoup
Professional
iVenture Accelerator Cohort 12 - Cozad New Venture Challenge Award Winner
- Launched Gemini tool-calling agent that parses client emails, maps logo design instructions to OpenCV/PIL image functions, and generates product mockups at scale, cutting order time from 2 days to <60s.
- Expanded tool coverage to 500+ product SKUs by building a Konva.js logo bounding box placement tool that cut onboarding time from 5 min to 3s per item.
- Enabled reliable deployment by containerizing the FastAPI backend on AWS with Docker, PostgreSQL, and S3, supporting both email and browser-based order submission to handle 35% more orders.
- Developed an end-to-end direct mail automation platform that identifies local businesses as sales targets using Playwright, BeautifulSoup, and Google Places API to generate 90,000+ enriched leads with a 93% success rate.
- Used GPT-image-1 to generate logo-matched backgrounds and automatically flag low-quality artwork for quality analysis review, increasing the number of orders handled by 15x.
- Deployed a serverless image enhancement service on AWS Lambda with a Real-ESRGAN ONNX model for 4x logo upscaling to process client logos before artwork generation.
- Shipped the full-stack platform in TypeScript, React, and PostgreSQL on AWS Lightsail, integrating TekWeld APIs for order fulfillment and an analytics dashboard for promo code and order tracking.
- Built automated Python reporting pipeline for a quantitative bond analytics platform, cutting generation from 10 min to 2s across 50-slide investor decks and PDFs.
- Architected visualization engine rendering 25+ stress curves and tornado plots with matplotlib, NumPy, and its gridspec module.
- Built CI/CD pipeline in GitHub Actions with Docker-containerized pytest, catching 4x more regressions pre-deploy on AWS.
- Built a fantasy football trade evaluation tool integrating Sleeper and ESPN APIs to compare 500+ player valuations in real time.
- Engineered XGBoost regression model achieving 22% higher trade valuation accuracy over flat point-total baselines on 3+ seasons of historical data; validated via k-fold cross-validation using 15+ engineered features.
Academic
GPA: 3.91 · Edmund J. James Scholar · Dean's List (SP25, FA25) · H&R Block 2025 CAPS Scholar
CS 128 · Intro to Computer Science II
CS 173 · Discrete Structures
CS 225 · Data Structures
CS 222 · Software Design Lab
CS 233 · Computer Architecture
CS 357 · Numerical Methods I
CS 374 · Intro to Algorithms
CS 411 · Database Systems
CS 443 · Reinforcement Learning
CS 341 · System Programming
CS 498 · Machine Learning Systems
MATH 241 · Calculus III
MATH 257 · Linear Algebra
STAT 107 · Data Science Discovery
STAT 400 · Statistics & Probability I
STAT 410 · Statistics & Probability II
STAT 425 · Statistical Modeling I
STAT 440 · Statistical Data Management
University of Illinois Urbana-Champaign
- Taught dynamic programming, induction, reductions, and graph modeling to 500+ students via tutorials and weekly office hours.
University of Illinois Urbana-Champaign
- Boosted exam performance by 15% (90% of course grade) by leading weekly review sessions for 250+ students.
- Selected as 1 of 3 lead CAs for summer cohort, improving problem comprehension 85% via structured in-class walkthroughs.
University of Illinois Urbana-Champaign
- Built a statistical analysis pipeline in Python to detect and quantify systematic AP Top 25 voter bias favoring high-profile NCAA conferences over comparable teams from smaller conferences for 250,000+ historical ballots.
- Isolated 10 most extreme voter–team bias pairs from 1,000+ candidates via custom Weighted Spearman footrule scoring pipeline.
- Surfaced bias patterns across 60+ schools with an interactive R Shiny geospatial heatmap for UIUC Research Symposium.
Projects
- Built a privacy-first fitness route planner that automatically reroutes around 31,000+ real-world ALPR surveillance cameras.
- Reduced camera load time by 5× by deploying on Cloudflare Workers, using edge runtime to cache geospatial queries and proxy API traffic without hitting the origin server.
- Enabled cross-device route sync by integrating Neon Postgres with Clerk OAuth, migrating localStorage routes to DB on first login.
CS-411 Database Systems
- Architected a full-stack relocation ranker, cutting research time 70% across 5,000+ U.S. cities via React, Flask REST API, and Cloud SQL on GCP.
- Increased ranking relevance 88% via Gemini LLM pipeline converting prompts to weights using MySQL procedures and triggers.
CS-222 Software Design Lab
- Built a browser-based graphic design editor with a layered canvas, supporting per-layer drag-to-reorder, visibility/lock toggles, and non-destructive transform (resize, rotate, translate) powered by react-konva.
- Engineered a tabbed multi-project workspace backed by Redux Toolkit, isolating canvas state (dimensions, zoom, background color) and layer stacks per project with paginated switching across unlimited sessions.
- Implemented pointer-anchored zoom from 10–500% with constrained Space+drag panning and fit-to-screen auto-scaling, alongside a tool sidebar (select, brush, shapes, text) for full canvas navigation and drawing control.
HackIllinois 2026
- Built a multi-agent adversarial legal simulation engine that spawns 3–5 AI judges with distinct legal philosophies, fans out GPU inference in parallel across Modal serverless workers, and streams a 3-round debate to the client over WebSocket.
- Automated jurisdictional research by chaining a Tavily web search with a GPT-4o-mini + Instructor structured-output call that extracts relevant legal precedents and demographic jury biases from any uploaded case PDF.
- Produced deterministic verdicts by routing the completed debate transcript through a GPT-4o Arbiter that scores argument quality, penalizes logical fallacies, and weights jury alignment by real jurisdictional bias data.
HackIllinois 2025
- Surfaced human-centered coding solutions from Reddit and StackOverflow by building a search engine that combines NLP, FAISS, and Gemini for semantic similarity retrieval.
- Built using Python, Flask, React, and Tailwind; integrated Reddit and StackExchange APIs; resolved semantic filtering challenges through vector-based similarity search.
Sandia Datathon
- Analyzed brainwave data with Python, pandas, and Scikit-Learn to model bilingualism.
- Visualized N400 ERP effects with Matplotlib and engineered repetition-based features to enhance classification.
Personal
outside of tech, I like to play sports (soccer, basketball), watch sports, go to the gym, play brawl stars, watch movies with my girlfriend, and travel!
Countries I've visited
States I've visited