CS honors grad who builds full-stack software, designs end-to-end data pipelines, and does embodied-AI research. I like problems that span the stack: turning raw data into a usable system, and a research idea into something that runs at scale.
I'm a developer and data engineer with embodied-AI research experience. I build full-stack software, design end-to-end data pipelines, and turn research ideas into things that run at scale, most recently a robotic-manipulation dataset pipeline in MSU's ACTION Lab.
My capstone, delivered to Henry Ford Innovations, was a full-stack healthcare platform: Angular 18 + Node/Express + Microsoft SQL Server, a 3-tier role hierarchy, and an audit-logging layer I owned end to end. Across my work the throughline is the same: take something ambiguous, make it concrete, ship it, measure it.
Right now I'm open to entry-level software engineering roles where I can keep working across the full stack, from data layer to interface.
Full-stack healthcare platform delivered to a real external partner, replacing a retired system for Henry Ford's Faith Community Nursing Program. Covers the whole encounter lifecycle: intake, visit recording, cost-savings tracking, certification management, and analytics. I owned the system-admin surface, the audit-logging layer, and major analytics features across a 3-tier RBAC model.
The data engine behind a robot-learning project: raw demos to a pixel-accurate segmentation dataset. I collected manipulation demos with a UMI gripper across 3 scenes and 16 task variations, annotated frame-0 boxes on 348 videos, then wrote a Python pipeline using SAM2 to propagate those prompts through entire videos, running the compute-heavy segmentation as batch jobs on MSU's HPC cluster.
A 40-year, 7,643-film study of what drives box-office success, then a model to predict it. Built an end-to-end analysis pipeline (correlation matrix, multi-decade trend viz, log-scale outlier handling), then adapted the SIR epidemiological model to simulate how a film's popularity spreads through an audience, solving the ODE system with SciPy and validating against 140+ days of real daily revenue.
Production mobile UI for a client's customer portal, plus an NFC tap-to-transfer prototype. I built a themed bottom-tab navigator across 9 screens with internationalized labels in 3 languages, resolved a macOS→Windows build blocker and onboarded 2 developers, then prototyped the NFC concept in React Native and validated it on physical hardware.
A collaborative-filtering recommender built solo on live API data, not a toy dataset. I owned every layer: ingesting 50 users' listening history via the Last.fm API, transforming 2,085+ unique tracks into structured DataFrames, building a sparse user-track matrix, computing a 50×50 cosine-similarity matrix, and writing a parameterized k-NN collaborative filter to surface unheard songs.
A serverless job-tracking service modeling HPC-style scheduling, with a persistent scheduler daemon. Built cloud-native on AWS: Lambda compute, API Gateway, DynamoDB persistence, and infrastructure defined in CDK, with CI/CD through GitHub Actions. Carries my HPC batch-scheduling background into a cloud-native design.
A cross-platform fitness tracker covering programs, progress logging, and analytics. Built mobile-first with a managed backend, shipping to both iOS and Android from a single codebase.
Co-authored an IEEE-published paper proposing an agentic AI framework for combat triage and mission adaptation, grounded in distributed cognition. It fuses multimodal data from drones, wearables, and battlefield sensors to assess injury severity, threat, and mission status in real time.
The system pairs survival-regression and triage-consistency models trained on large-scale EMS data (NEMSIS) with reinforcement-learning agents that simulate UAV coordination, evacuation timing, and dynamic route adaptation under adversarial conditions, surfacing explainable recommendations for medics, pilots, and commanders.
In MSU's ACTION Lab, I built the data engine behind a robotic-manipulation learning effort. I wrote a Python pipeline using SAM2 to propagate frame-0 annotations through full videos, running segmentation as batch jobs on the HPC cluster.
The dataset feeds the lab's downstream model training: the precise, large-scale ground truth that hand-labeling couldn't produce.
Open to entry-level software engineering roles and research collaborations. The fastest way to reach me is email.