[ ]
I’m a Computer Science student at Purdue building machine learning systems for the defense world — OSINT ingestion, hybrid retrieval, and decision-support models that hold up in production.
This is my little corner of the internet — projects, experiments, research, and whatever I’m learning along the way.
US CitizenClearanceWest Lafayette, INOpen to Summer '27
0.00
GPA / Purdue CS
0M+
data points modeled
0K+
documents indexed
0×
LLM cost reduction
0
unlabeled images trained
/about_me
I’m an ML Systems Engineer Intern at a stealth defense market-intelligence startup in Quantico, where I build OSINT ingestion and hybrid retrieval that feeds an LLM judge.
Before that I was an MLOps intern at U.S. Marine Corps Systems Command, taking decision-support models from prototype to production on Databricks — and before that, streaming live race telemetry through Kafka and Flink for a World Racing League team.
I’m in my junior year of a B.S. in Computer Science at Purdue. On the side I train self-supervised energy models that judge whether an image actually holds together.
The stack I reach for:
languages
PythonJavaCRSQL
ml / ai
PyTorchTensorFlowKerasHugging Face Transformersscikit-learnXGBoostLightGBMSHAPMLflow
data
pandasNumPyMatplotlibPySparkDatabricksKafkaFlinkApache IcebergPostgreSQLpgvector
platform
AWS BedrockAWS ECSTerraformDockerFastAPIGitLinuxOpenTelemetryAgile
focus
LLMsRAGNLPMLOpsSelf-Supervised Learning
Outside of work, you’ll usually find me hiking, playing anything that involves a racket or paddle, and catching Steelers games. Oh! I'm also an avid mixologist.
/experiences
ML Systems Engineer Intern (CUI)
Stealth Startup | Defense Market Intelligence SaaS
activeJuly 2026 — Present · Quantico, VA
Architected OSINT ingestion over 50+ RSS/GDELT sources with an SSRF-safe fetcher, MinHash dedup, and 8K+ article clustering
Built pgvector + BM25 hybrid retrieval with MiniLM embeddings and tenant-scoped dismissals feeding a gpt-oss-20b LLM judge
Cut LLM profile ingestion cost to 1/35th by migrating to AWS Bedrock; shipped token telemetry and CI golden-set evals
Deployed FastAPI and PostgreSQL services on AWS ECS via Terraform with SQLAlchemy and Alembic schema migrations
/projects
JEPA Coherence Model
Self-supervised energy-based model that scores whether an image is internally coherent, with no labels at training time.
selected
Ames Housing Price Analysis
Statistical and ML analysis of how basement, first-floor, and second-floor square footage differently impact home prices.
click
Radar Signal Processing Simulator
Radar simulation demonstrating signal processing pipelines, target detection algorithms, and multi-target tracking with Kalman filtering.
click
Java Chat Application
Real-time client-server messaging system with Swing GUI and optional MySQL persistence.
click
JEPA Coherence Model
Self-supervised energy-based model that scores whether an image is internally coherent, with no labels at training time.
Trained a joint-embedding predictive architecture in PyTorch on 7,420 unlabeled images; incoherent inputs separate at 0.70 AUROC.
Ran multi-seed ablations against MAE, contrastive, and PMI baselines — training duration mattered more than model capacity.
Frozen encoder transferred to label-scarce classification and beat from-scratch training in all 12 evaluation cells.
Set Transformer encoders for set-level anomaly detection; full scaling study tracked in Weights & Biases.
/credentials
Purdue University
B.S. Computer Science
graduationMay 2028
gpa3.78 / 4.00
locationWest Lafayette, IN
clearanceClearance · US Citizen
relevant coursework
Data Mining and Machine LearningApplied Regression AnalysisProbability and StatisticsData Structures and AlgorithmsLinear AlgebraOperating SystemsProgramming in CObject-Oriented ProgrammingRelational DatabasesCompetitive Programming
awards · leadership · licenses
Purdue Presidential Scholarship
Merit award, Purdue University
Clearance
US Citizen · Department of War sponsored
Computer Vision Target Localization — Team Lead
Purdue National Security and Defense Society
Teaching Assistant
Data Structures and Algorithms · Programming in C
FCC Amateur Radio Technician License
Licensed operator
/film_strip
Purdue x Google AI Summit