Clarity over cleverness
Code should explain itself. If something needs a paragraph of comments, it usually wants to be simpler.
Open to internships & research opportunities
Colten Hargett, Computer Science and Data Science student building intelligent software
PythonMachine Learningscikit-learnpandasNumPyJavaLLM AgentsRAG PipelinesPythonMachine Learningscikit-learnpandasNumPyJavaLLM AgentsRAG Pipelines
ChromaDBGemini APILangChainBacktestingData ScienceAutomationHTML / CSSUnixChromaDBGemini APILangChainBacktestingData ScienceAutomationHTML / CSSUnix
01About
I'm a Computer Science and Data Science student who loves building things that take messy, real-world data and turn it into something genuinely useful: forecasting models that explain themselves, AI agents that do the reading for you, and software that feels considered down to the details.
Glen Allen, VA
→ Baltimore, MD
Class of '29
A bit more about me
I grew up in Glen Allen, Virginia, just outside Richmond, and I'm now in Baltimore studying Computer Science and Data Science at Loyola University Maryland, where I'm part of the Hyman Science Scholars Program.
I like the moment when a pile of raw data turns into something you can actually use. That's why I paired CS with data science, and it's what most of my projects chase: a forecasting model that can explain its own predictions, or an agent that reads the day's news so you don't have to.
Before I was writing models, I spent three summers managing a pool and leading a staff of 30+ lifeguards. It taught me to stay calm under pressure, own problems from start to finish, and make sure the people around me have what they need. I bring that same mindset to software.
Off the clock, I've volunteered with James River Greyhounds since 2020, I co-founded my high school's Musical History Club, and I usually have one eye on the markets.
- Based in
- Baltimore, MD
- From
- Glen Allen, VA
- Studying
- B.S. CS + B.S. Data Science · '29
- Interested in
- Machine learning, data tools, automation
- Looking for
- Internships & research
- Off the clock
- Greyhound rescue, music history, markets
0
Projects built
0+
People led
0
Majors
0.0
HS weighted GPA
How I build
Four principles that show up in every project, from a Java console app to a machine-learning pipeline.
Built for messy reality
Real data is never clean. I validate inputs and design for edge cases so things keep working when conditions aren't ideal.
Measure, don't guess
Every model gets a baseline and a backtest. If it can't beat the simple answer, it isn't done yet.
AI where it earns its place
I reach for machine learning when it produces measurable value, not to check a box.
Toolbox
What I reach for day to day
Languages
- Python
- Java
- HTML / CSS
- Unix shell
Data & ML
- pandas
- NumPy
- scikit-learn
- yfinance
AI systems
- Gemini API
- LangChain
- ChromaDB
- RAG
Workflow
- Git & GitHub
- PyCharm
- Automation
- Backtesting
02Selected work
Systems thatthink with data.
Two projects I'm proudest of, each with a live, interactive look at how it works under the hood.
01Machine Learning · Forecasting
Stock Market Predictor
Live · close of Sep 23, 2026
Next-session high forecast
$337.78
+0.22% vs. $337.02 close, weighted from 5 similar days
Walk-forward backtest
1,133 predictions
k-NN model1.04%
Baseline: today's close1.28%
Baseline: today's high1.13%
Mean absolute % error on next-day highs. Lower than both baselines.
Predicted vs actual high · last 40 sessions
Actual PredictedJul 28 – Sep 22
Most similar days in history
- k1Mar 27, 202325%+0.13%
- k2Jan 26, 202422%-0.11%
- k3Oct 13, 202319%+0.13%
- k4May 18, 202618%+0.90%
- k5Dec 3, 202517%+0.20%
Weight · how high the stock went the next day
Real daily prices from Yahoo Finance · model re-run every 6 hours · for demonstration, not investment advice
A similarity-based forecasting engine that predicts a stock's next-day high by finding the moments in history that looked most like today.
The problem
Most market models are black boxes. I wanted one I could actually reason about: every prediction explained by the real historical days behind it.
How it works
- 01Engineered 11 features per trading day: returns, spreads, volume change, moving-average gaps and volatility
- 02Standardized feature space + k-nearest-neighbors to find the most similar historical days
- 03Inverse-distance weighting, so the closest matches carry the most influence
- 04Walk-forward backtesting that only ever sees the past, benchmarked against two naive baselines
1.63%
Avg. error, live backtest
4 / 4
Tickers beating both baselines
4,532
Walk-forward predictions
Pythonscikit-learnpandasNumPyyfinance
02AI Agents · Automation
News Summary Agent
Last run · 4:10 PM ET79 articles · 6 stories retrieved
79 articles · 6 stories retrieved
Inbox · 4:10 PM ETTop stories
Most-covered stories, last 24 hours
- Harvey Weinstein sentenced to 15 years in prison for sexual assaultBBC · CBS · NBC
- Police make 2nd arrest in death of woman found hanging from treeCBS · NBC
- Iran's president gives a rare wartime speech at the U.N.NPR · BBC
- What is the White House press pool and how does it work?CBS · NBC
- FBI investigating hacking group’s claim of massive breach of agent infoNBC
Live · the site re-runs this pipeline every 6 hours on the real feeds
An autonomous pipeline that reads the day's news from five major outlets, indexes it into a vector database, and delivers a newspaper-style briefing by email every night.
The problem
Keeping up with the news takes time and a dozen tabs. I wanted a system that does the reading for me and hands back one clean, trustworthy summary.
How it works
- 01RSS scraper pulls the last 24 hours from NPR, BBC, ABC, CBS and NBC and extracts full article text
- 02Articles are chunked with LangChain and embedded into a persistent ChromaDB collection
- 03Gemini writes a grounded, editor-style recap using only the retrieved context
- 04A scheduler runs the full pipeline nightly and emails the briefing out automatically
79
Articles read, last 24h
5
News sources
0
Manual steps
PythonGeminiChromaDBLangChainRSS
03Archive
EverythingI've built.
No.ProjectDescriptionStackCode
- 01Stock Market Predictork-NN forecasting of next-day highs with walk-forward backtesting.
- 02News Summary AgentScrape → embed → summarize → email, fully automated every night.
- 03Movie AnalyzerParses movie datasets and produces structured, queryable output.
- 04Restaurant AnalyzerReads restaurant records from files and surfaces useful insights.
- 05Song AnalyzerProcesses song and artist catalogs with custom analysis functions.
- 06Morse Code TranslatorBidirectional text ↔ Morse translation, including file input.
- 07Dice Rolling SimulatorSimulates rolls and visualizes how outcomes distribute over time.
- 08ASCII Art MakerTurns plain text input into generated ASCII artwork.
- 09HangmanCommand-line hangman with thorough input validation.
- 10Rock, Paper, Scissors+A five-choice strategy variant with custom rules and outcomes.
- 11ATM SimulationAccount balances, transactions and input checking in the console.
- 12Grocery List MakerBuild, edit and manage grocery lists interactively.
- 13Personality TestAn interactive, scored personality quiz for the command line.
- 14Java ExercisesEight problem sets drilling core language fundamentals.
04Journey
Learning,leading,building.
The classrooms, pool decks and volunteer events that shaped how I learn, lead and build. Leadership came first, and it still shapes how I work on every project.
- 2025 — 2029Education
B.S. Computer Science & B.S. Data Science
Loyola University Maryland · Baltimore, MD
Double-majoring at the intersection of software engineering and data. Building ML and automation projects alongside coursework.
- Hyman Science Scholars Program
- Alpha Kappa Psi
- Information Systems Student Organization
- Summer 2025Leadership
Head Guard
Coastline Aquatics · Glen Allen, VA
Supervised lifeguard teams, ran onboarding and training, and designed follow-up processes that cut down on recurring issues.
, - 2022 — 2024Leadership
Pool Manager
SwimMetro Management · Glen Allen, VA
Led 30+ lifeguards across three seasons: scheduling, training, patron events and incident response alongside first responders.
, - 2021 — 2025Education
Advanced Diploma · 4.3 Weighted GPA
Deep Run High School · Glen Allen, VA
Co-founded and served as Vice President of the Musical History Club. Member of FBLA and the Finance & Investment Club.
, - 2020 — PresentCommunity
Volunteer
James River Greyhounds · Richmond, VA
Help plan bi-annual raffles averaging $2,500 raised and work directly with the organization's president on adoption events.
05Contact
Let's buildsomething great.
I'm looking for internships, research and project collaborations in software engineering, data science and machine learning. The fastest way to reach me is email. I usually reply within a day.