Hello, allow me to introduce myself...

I'm Richard Roberson, a first-year graduate student pursuing a Master's in Computer Science at the University of Colorado, Boulder, where I also completed my undergraduate degree in Computer Science. My journey into technology began in middle school with iOS app development, sparking a lasting interest in coding and problem-solving. I bring professional experience from Ricoh, where I developed enterprise software solutions using modern technologies like Docker, Kubernetes, and cloud platforms. Currently, I focus on AI and machine learning, specializing in LangChain, LangGraph, and AI-driven applications, alongside research including a paper on text-to-SQL conversions. I'm a driven learner, excited to take on new challenges, expand my skills, and contribute meaningfully to the field of technology.

University of Colorado Boulder College of Engineering

Bachelor's of Science in Computer Science

2020-2024

University of Colorado Boulder College of Engineering

Master's in Computer Science

2024-Present

Work Experience

Machine Learning Research

@ Programming Languages and Verification at the University of Colorado Boulder

June 2023 - February 2024

Leading a research project that finetunes large language models for text-to-SQL synthesis. Learn more about the project in the Recent Projects section below.

Python

C++

LLM Finetuning

Fullstack Freelance Developer

June 2024 - May 2025

"Richard was able to quickly complete the tasks assigned to him, along with finding and fixing a few issues he found on his own. He quickly was able to understand what was being asked and was able to take initiative and perform his tasks despite our somewhat half-baked and incomplete requirements. He will be one of the first places we look for front-end development work from now on." - Client Feedback

Frontend Development

Backend Development

AI Software Engineer Intern

@ Ricoh, USA in Boulder

May 2025 - August 2025

Developed a natural language SQL database interaction chatbot that uses graphs and plots. Enables non-technical users to interact with company databases using natural language queries. in its outputs for internal use and the infrastructure required to host it on-premise.

Pivoted to build a production ready version as an agent deployed using AWS Bedrock AgentCore services, which enables background running agents and great observability for enterprise use cases.

LLMs

Text-2-SQL

AI Agents

Fullstack Development

Recent Projects

NATS JetStream AI Pipeline w/ Dashboard

Kubernetes

GKE

NATS

JetStream

Docker

Next.js

Sentiment Analysis

Topic Classifaction

Microservices

Cloud

A complete data processing pipeline using NATS JetStream for messaging between microservices, with support for cloud-native Kubernetes deployment on Google Kubernetes Engine (GKE). Features cost-optimized GKE cluster management, real-time sentiment analysis, and a Next.js visualization dashboard. Includes NATS server with JetStream persistence, firehose data ingestion, stream processing with sentiment analysis, and mock data generation for development.

Visit

K8s Agents Service

Kubernetes

LangGraph

FastAPI

Streamlit

Next.js

YugabyteDB

RAG

ArgoCD

Helm

Docker

GitOps

LLMs

A production-ready Kubernetes deployment platform for AI agents built with LangGraph, FastAPI, and Streamlit. Features comprehensive Helm charts with dual-architecture deployment, ArgoCD GitOps integration, YugabyteDB with pgvector for RAG capabilities, and automated cron jobs for vector database maintenance. Includes both Streamlit and Next.js UIs with multi-agent support for resume assistance and web research.

Visit

Analyzing the Effectiveness of Large Language Models on Text-to-SQL Synthesis

Research

LLM

Machine Learning

Python

Investigating various approaches to using Large Language Models (LLMs) for Text-to-SQL program synthesis, focusing on the outcomes and insights derived. Employing the popular Text-to-SQL dataset, spider, the goal was to input a natural language question along with the database schema and output the correct SQL SELECT query.

Visit

Research Podcasts Platform

Kubernetes

Kafka

RAG

YugabyteDB

TTS

Flask

React

TanStack

arXiv

LLMs

Azure OpenAI

A distributed research analysis system using RAG to process academic papers from arXiv and generate AI-powered podcast summaries. Features event-driven architecture with Kafka for reliability, YugabyteDB for distributed SQL storage with pgvector, high-quality TTS audio generation, and real-time progress tracking via server-sent events. Includes semantic similarity-based recommendations and comprehensive Kubernetes deployment.

Visit

Custom Bluesky feed in a home lab environment

Bluesky

ATProto

Kubernetes

YugabyteDB

Flask

Python

Cloudflare

NGINX

Home Lab

Brandon Sanderson

A specialized ATProto feed service for Brandon Sanderson's Cosmere universe fans. Features intelligent filtering using comprehensive keyword sets, trending and chronological post combination, Kubernetes microservices architecture with YugabyteDB, and home lab deployment with SSL/TLS termination. Includes web API service, firehose processing, scheduler service, and automatic DNS management with Cloudflare.

Visit

Web-based text-to-speech document reader

Next.js

TTS

PWA

IndexedDB

Docker

Tailwind CSS

Kokoro

OpenAI

EPUB

PDF

Audiobooks

A comprehensive document reader with TTS capabilities supporting EPUB, PDF, TXT, MD, and DOCX formats. Features real-time read-along experience with highlighted text, audiobook creation and export (M4B format), mobile PWA support, customizable TTS settings, and local-first architecture using IndexedDB. Compatible with multiple TTS APIs including Kokoro-FastAPI, Orpheus-FastAPI, and OpenAI.

Visit

Using Large Language Models for Text-to-SQL Synthesis

Research

Text-2-SQL

Finetuning

LLMs

This project fine-tunes a local large language model (LLM) WizardCoder-15B with QLoRA on a custom Spider training dataset to create a Text-to-SQL LLM. The resultant model, achieves 61% execution accuracy, incorporating database context for validation.

Visit

TuneTidy iOS App

React Native

iOS

Spotify API

TanStack Query

Firebase

OpenAI

DALL-E

OAuth

Analytics

A React Native iOS app that enhances Spotify with advanced playlist management and analytics. Features comprehensive playlist sorting by audio features (danceability, energy, tempo, etc.), top tracks/artists/albums analysis across multiple time ranges, AI-generated playlist cover art using DALL-E, real-time playlist reordering with progress tracking, and OAuth 2.0 authentication with PKCE.

Visit

TuneTidy Web App

React.js

Spotify API

Chakra UI

OAuth

Netlify

Music Analytics

Playlist Management

A comprehensive React.js web application for Spotify playlist management and music analytics. Features advanced playlist sorting by multiple criteria including audio features (acousticness, danceability, energy, etc.), listening statistics with time-based analysis, playlist editing capabilities, and user authentication via Spotify OAuth. Built with Chakra UI and deployed on Netlify.

Visit

Artificial Intelligence Research

Research

LLMs

GANs

Written in 2021 (before the AI craze). A research paper posing the question: Why are Generative Adversarial Networks (GANs) Important for the Advancement of Artificial Intelligence?

Visit

Manacala Game GUI Application

C++

Created with Qt Creator using C++.

Visit

Shoots and Ladders GUI Application

C++

Created with Qt Creator using C++.

Visit

Skills

Favorites

Python

Expert

FastAPI

Intermediate

PostgreSQL

Proficient

Kubernetes

Proficient

Contact Info

Send me an email directly from this page or text me. I will answer as soon as possible

richard.roberson@colorado.edu

Other Email

richardr1126@gmail.com

Phone

(561) 267-3440

Location

Boulder, CO

Richard Roberson's Portfolio

Work Experience

Machine Learning Research

@ Programming Languages and Verification at the University of Colorado Boulder

Fullstack Freelance Developer

AI Software Engineer Intern

@ Ricoh, USA in Boulder

Recent Projects

NATS JetStream AI Pipeline w/ Dashboard

K8s Agents Service

Analyzing the Effectiveness of Large Language Models on Text-to-SQL Synthesis

Research Podcasts Platform

Custom Bluesky feed in a home lab environment

Web-based text-to-speech document reader

Using Large Language Models for Text-to-SQL Synthesis

TuneTidy iOS App

TuneTidy Web App

Artificial Intelligence Research

Manacala Game GUI Application

Shoots and Ladders GUI Application

Skills

Favorites

Programming Languages

Tools & Frameworks

Databases & Messaging

DevOps

Contact Info

Resume