· PORTFOLIO MMXXVI
Knoxville, Tennessee · University of Tennessee

Devin
Colby.

Applied AI engineer
& historian — building
systems where language
models meet human context.

B.S. APPLIED ARTIFICIAL INTELLIGENCE
B.A. HISTORY
UNIVERSITY OF TENNESSEE, KNOXVILLE
EXPECTED GRADUATION · DEC. 2026
§ I  —  About

I'm a final-year double-major at the University of Tennessee, Knoxville, working at an intersection most people don't expect: applied artificial intelligence and history. Most of what I build involves large language models — a deployed full-stack PWA for shared movie recommendations, a RAG system for academic advising, an offline AI that runs on consumer hardware. But the historical training matters as much as the engineering does; it's what teaches me to ask better questions about what we're building, and why.

❦
§ II  —  Selected Work
03 PROJECTS
PROJECT  01
Live

WatchWise ↗

A shared recommendation engine for couples & friends.

A full-stack Next.js Progressive Web App that lets users and their partners track and rate movies and shows together, using a four-way rating system to surface personalized recommendations powered by the Anthropic API. Built end-to-end including authentication, persistent storage with Supabase, and sync between paired accounts.

Next.js Supabase Anthropic API Vercel PWA
View Live ↗ Source ↗
PROJECT  02
Working Prototype

Advising Vols

RAG-powered academic advising for the University of Tennessee.

A Retrieval-Augmented Generation system that lets students and academic advisors query UTK policy, course, and enrollment data in natural language — replacing manual lookups across scattered PDFs and portals. Includes a custom PDF parsing pipeline that ingests DARS (Degree Audit Reporting System) documents across multiple program formats, plus separate student- and advisor-facing portals. Built with a five-person team.

Python RAG LangChain Vector DB Streamlit PDF Parsing
Source ↗
PROJECT  03
Completed

Offline AI System

Local LLM inference, no internet required.

A fully offline AI system that runs large language models locally on consumer hardware — no cloud, no API calls, no internet dependency. Implements a local RAG pipeline with ChromaDB for vector storage, supporting document-grounded question answering with low-latency private inference. Includes formal HCI documentation and Wizard of Oz usability testing.

Ollama Phi-3 Mini ChromaDB Python Local RAG
Source ↗
§ III  —  Toolkit
LANGUAGES & FRAMEWORKS

Python · JavaScript · React · Next.js

AI & MACHINE LEARNING

Retrieval-Augmented Generation · LangChain · Hugging Face · Ollama · ChromaDB · Anthropic API · Prompt engineering · Local LLM deployment

DATA & BACKEND

pandas · scikit-learn · NumPy · REST APIs · FastAPI · Flask · Supabase · Vercel

TOOLS & VISUALIZATION

Git · GitHub · Tableau · PDF parsing pipelines · Data ingestion

HARDWARE & SYSTEMS

Computer hardware troubleshooting · CompTIA A+ (in progress)

§ IV  —  Background
RECENT
East Tennessee Historical Society & Museum
ARCHIVAL RESEARCH · KNOXVILLE, TN

Supported archival research and public programming, organizing and cataloging large volumes of primary source materials and contextualizing documents for research and public-facing exhibitions.

PRIOR
Howard Baker Center for Public Policy
OUTREACH & EVENTS · UTK

Assisted with event coordination and public outreach for a nonpartisan public policy center, communicating across politically diverse stakeholder groups.

PRIOR
Whole Foods Market
CUSTOMER SERVICE · KNOXVILLE, TN

High-volume retail. Reliability, teamwork, and operating under pressure.

§ V  —  Get in Touch

Currently
seeking roles
in AI, data,
& software.

Most interested in roles where I can ship real systems — applied AI, LLM tooling, data engineering, or product-focused software work. Reach out for full-time opportunities, collaborations, or just to talk shop.

LOCATION
Knoxville, Tennessee