From mechanical engineering to human-centered AI
I recently completed my PhD in Computer Science at the University of Oklahoma, advised by Dr. Chris Weaver. My dissertation developed topology-centric visual analytics — new representations and interactions that help people explore, compare, and reason about complex hierarchical data, going far beyond simple parent-child tree views.
Today I'm an AI Software Engineer at Citi, where I build production LLM platforms: GraphRAG knowledge discovery over siloed technical knowledge, retrieval and evaluation frameworks that catch quality regressions before users do, and analytics that turn model behavior into decisions product teams can act on.
The thread connecting all of it: AI is only useful when humans can verify it. Whether it's a mammogram classifier, a RAG pipeline, or a tree with ten thousand nodes, I design systems that show their evidence — combining research rigor in HCI and visualization with the engineering to ship at enterprise scale.
Education
Certifications
Where I've built things
From oil rigs in Qatar to enterprise LLM platforms — a decade of turning messy real-world data into decisions.
- Built a production LLM analytics platform on Vertex AI with FAISS-backed retrieval, cutting retrieval latency by 40% across enterprise workflows.
- Designed a GraphRAG knowledge discovery system using semantic triples and vector embeddings for grounded multi-hop reasoning across siloed technical knowledge.
- Created an LLM evaluation framework — gold-standard mapping, regression tests, retrieval checks — that surfaces failure clusters and quality regressions for product teams.
- Contributed to an LLM-assisted migration of 200K+ lines of legacy JavaScript/XML to React micro-frontends, reducing manual effort by 60%.
- Built an Angular visual authoring platform replacing manual XML editing with canvas-based flow mapping and AI-assisted authoring; +40% productivity.
- Created Hieros, a visualization system for representing, coordinating, and interacting with composite topologies in hierarchical data.
- Developed C4D3, a TypeScript coordination library for D3.js powering 30+ synchronized views over 10,000+ data points.
- Fine-tuned a 328M-parameter foundation model and a 36M Vision Transformer for multi-view mammogram cancer prediction — the compact model won at 80.4% accuracy.
- Engineered a multi-user VR learning platform with REST APIs synchronizing fNIRS, eye-tracking, and haptic biometric streams in real time.
- Taught Intro to Python (CS1213) as instructor from 2020 onward; conducted eye-tracking studies on visual attention in media tasks.
- Built SARIMAX and gradient-boosting forecasting pipelines behind a Flask API, improving KPI forecast accuracy by 12%.
- Automated SQL/Python analytics for engagement metrics, turning stakeholder questions into repeatable reporting workflows.
- Built Python, SQL & SAS workflows over maintenance, inventory, and procurement data for oil & gas operations — contributing to a 15% reduction in unplanned outages.
- Standardized Bill of Materials, SPIR validation, and equipment hierarchies across engineering and supply-chain teams.
- Automated offshore resource planning and logistics reporting with Python, Excel macros, and SAS extraction tools.
Selected research & engineering work
Systems for coordinating views, understanding trees, and grounding AI — click any card to dive in.
C4D3 — Coordination for Data-Driven Documents
A view-level abstraction and coordination library for D3.js that makes building coordinated multiple views declarative — 30+ synchronized visualizations, 10,000+ points. Published at IVAPP 2025.
Explore the live demos
Representational Suitability of Tree Visualizations
When does a treemap beat a node-link diagram? A model and evaluation framework for matching hierarchical visualization techniques to data topology and task. IEEE VIS 2022 poster.
Read the research
Agentic AI, RAG & LLM Applications
LangGraph agent workflows, local RAG pipelines, an LLM-powered skin-condition search app, and reinforcement learning experiments — hands-on generative AI systems.
See the projects
David vs. Goliath — Medical Imaging AI
Showed that a compact 36M Vision Transformer outperforms a 328M foundation model on multi-view mammogram breast cancer prediction (80.4% accuracy). Best Paper 2nd Place, SPIE 2024.
Read the paper
Gesture-Based Data Manipulation
Interactive gesture recognition for manipulating and exploring data visualizations — natural, embodied interfaces for human-data interaction.
See it in action
Interactive Visualization Portfolio
A gallery of hand-built D3 visualizations — COVID-19 explorers, animated ion simulations, coordinated dashboards, and more experiments in interactive data storytelling.
Browse the gallery
Vision Transformer with PEFT — MLOps Pipeline
End-to-end MLOps pipeline training ViTs for CIFAR-10 and multi-view classification with parameter-efficient fine-tuning — Dockerized, CI/CD via GitHub Actions, fully reproducible.
See the pipeline
GAN Image Generation — FastAPI Deployment
A GAN that dreams up handwritten digits, paired with a classifier that judges them — served through a FastAPI backend and web UI as a full-stack generative AI deployment.
Generate some digits
Bidirectional English ⇄ Sanskrit Transformer
One Transformer, two directions: translation between English and Sanskrit using direction tokens and a unified vocabulary, with custom large-file data loaders and a CI pipeline.
Read the codeResearch output
Peer-reviewed
Dissertation & manuscripts
My toolkit
LLM Systems & Agentic AI
Machine Learning & Deep Learning
Visualization & Frontend
Data, Cloud & MLOps
Awards, service & leadership
Awards & Honors
- Best Paper Award, 2nd Place — SPIE Medical Imaging (Computer-Aided Diagnosis)2024
- State of Outreach Innovation Award — OU College of Continuing Education2024
- CS Alumni Graduate Fellowship — University of Oklahoma2022, 2024
- Richard L. O'Shields Engineering Scholarship2023
- PhD Recruitment Excellence Fellowship — OU College of Engineering2018
- Academic Excellence Award — Army ROTC, University of Oklahoma2017
Academic Service
- Instructor — CS1213 Intro to Python, University of Oklahoma2020+
- Reviewer — IEEE VIS, EuroVis2023
- Student Volunteer — IEEE VIS2022
Leadership
- Graduate Student Community, OU — advocacy, inclusivity & funding initiatives
- Data Science & Analytics Association, OU — talks and community events
- Former Cadet, OU Army ROTC — blood drives, charity marathons, veteran causes