A. BIUCKIANS · DATA SCIENCE
M.S. Data Science · GWU · open to new-grad roles · May 2027

I model the question most models skip.

A classifier tells you what will happen. My work targets the harder question underneath it — whose decision can I actually change, and how few observations before I know for sure.

causal uplift modeling · adversarial detection · time-series & MLOps · deployed APIs & dashboards

a model that beats random
random targeting (the baseline to beat)
01

The angle I work from

Prediction is table stakes. The projects I'm proudest of reframe the problem so the answer is something a business can act on — and stay honest about what the data can and can't prove.

The usual approach

Rank everyone by risk score.

A churn classifier flags who will leave. A fraud model flags what looks anomalous. Useful — but it spends budget on people who'd stay anyway, and treats "flagged" as if it were "provably true."

predict(x) → probability
→ act on the top of the list
How I frame it

Rank everyone by whether I can move them — and how fast I'll know.

I estimate the causal effect of an intervention per person, route budget only to the persuadable, and quantify detection latency under an opponent who's actively hiding. Then I report the honest number, not the flattering one.

effect(x | treated) − effect(x | not)
→ act only where it changes the outcome
02

Flagship projects

Three end-to-end builds on real, messy data — a spatial model whose residuals recover a finding it was never told about, an operations system that goes from forecast to optimization to simulation, and the production monitoring that catches what drift detectors miss.

03

More projects

A run of explainable, decision-support ML — each one deployed as an interactive Streamlit tool, each pairing predictions with the reasoning behind them.

04

Toolkit

The stack behind the work — weighted toward Python ML, causal inference, explainability, and getting models out of the notebook and onto a URL.

Languages
  • Python
  • SQL
  • R
  • SAS
  • C++
  • JavaScript
ML & Data Science
  • scikit-learn
  • LightGBM · XGBoost
  • PyTorch
  • causalml · econml
  • SHAP
  • Optuna · SMOTE
  • LangChain · RAG
  • ChromaDB · embeddings
  • Ollama · local LLMs
  • pandas · NumPy · SciPy
Modeling
  • Uplift / CATE
  • Causal inference
  • Retrieval-augmented generation
  • Semantic search
  • Sequential testing (SPRT/CUSUM)
  • Anomaly detection
  • Time-series forecasting & CV
  • Calibration & fairness
MLOps & Serving
  • FastAPI · Pydantic
  • Docker
  • MLflow
  • GitHub Actions (CI)
  • pytest · ruff
  • Render · REST APIs
Visualize
  • Streamlit · Dash
  • Plotly · Matplotlib
  • Power BI · Tableau
  • seaborn · ggplot2
  • Shiny / Leaflet
05

Background

Experience

Jan – Apr 2024
Analytics Intern — D-I Baseball
High Point University

Analyzed Trackman pitch and batted-ball data for 30+ players to support evaluation; assessed team and opponent tendencies through Synergy and scouting workflows.

Aug 2022 – Nov 2023
D-I Basketball Team Manager
High Point University

Ran film analysis in HD Intelligence to identify high-efficiency shot zones.

Aug – Sep 2023
Film & Analytics Cohort
Grow the Game · Hudl

Generated box scores and shot charts for 40+ players; produced five solo film segments coding in SportsCode.

May – Jul 2023
Game Day Intern
Winston-Salem Dash

Tracked live gameplay with DakStats across 20+ games.

Education

M.S., Data Science
George Washington University, Washington DC
Expected May 2027 · GPA 3.72
B.A., Sport Management
Minor, Data Analytics · High Point University
Dec 2024 · GPA 3.55

Graduate coursework spans data mining, machine learning, deep learning, NLP, time-series modeling, and visualization of complex data.

06

Get in touch

Open to new-grad data science roles · May 2027

Let's find the customers worth moving.

Looking for new-grad data science and ML roles — especially anything applied, causal, or decision-support. Happy to walk through any project in depth, code and all.

Open to opportunities

SeekingNew-grad data science & ML roles
AvailableMay 2027, on graduation
Based inRockville, MD — open to remote, hybrid, or on-site
DegreeM.S. Data Science, George Washington University
FocusCausal / uplift modeling, adversarial detection, explainable ML