MAchine Learning and AI

Master's Project / Dissertation

  • Industry supported by L-Acoustics, in progress until August 2026. 
  • Crowdsourcing a captioned dataset of Room Impulse Responses to explore the human perception of acoustics in natural language. 
  • This research aims to provide data for better fine-tuning and training of automated audio captioning (AAC) models.
  • Investigate for any links between the parameters of Impulse Response (RT60, DRR, etc.) and human language. 

Highlighted projects

Beat Detection using TCN's

Re-implemented a SOA Beat Detection system and trained it on the Ballroom dataset, achieved ~90% accuracy in beat detection, and extended the system to detect downbeats. Built a spectral analysis dataset pipeline to efficiently train the model.

Emotion AND Gender Classification

Used a custom TCN to learn emotion and gender from VGGish embeddings of the RAVDESS emotion datasets spectral features. Evaluated both models in isolation and together. Had to calculate RF to ensure Classifier properly analyzes VGG embeddings.

Beat Generation with GAN's

Created a novel interactive system utilizing a GAN to make an interactive beat generator, with multiple genre selections. Performed a small qualitative evaluation study on beats to determine the systems metrical quality and accruracy in capturing genre.
  • Other Research
  • Live Sound
  • Post-Production
  • Published Writing
<
>

Other AI Projects

CIFAR-100 Multi-Modal Classifier
  • A Python based Multi-Modal planning and CLAP project. Uses multiple techniques to grow a vocabulary corpus of embeddings and use contrastive learning to train an Image classifier for a basic PDDL planner. 
  • Based on a series of coursework labs, the project starts with the Visual Genome corpus, and creates an embedding space using a Skipgram model. From there, the CIFAR-100 vocabulary is added into the embedding space using Genetic Algorithms to train an image classifier using contrastive learning. 
Audio Identification
  • Basic Python implementation of the original Shazam algorithm. Hashing a constellation matrix to create temporal and frequency invariant documents for a Query-by-Example (QBE) system. 
  • Evaluated the performance of different spectral features via MFCC, Mel/Log, and CQT with different windowing and hop parameters. Empirically measured the resulting constellation map densities to find the balance of robustness and speed using a hashing algorithm. 
​Dungeons & Dragons Ontology 
  • An OWL/XML Ontology that fuses semantic data from Wikidata, D&D Wiki, and dbpedia to find Hero's and Villains and associate individuals in the ontology with their 3e/5e character class.. 
  • Had to solve sparse data, find new relationships and equivalencies in the ontology through Description Logic and fuse data of individual characters from popular R.A. Salvatore novels with 3e/5e skills, saves, and equipment.

Live Sound Engineering

Sound Designer, Powerhouse Theatre and the New Canaan Town Players; 2023 - 2024
  • Edited and recorded sound effects for a tech heavy stage play.
  • Set up a sound reinforcement system.
Director of Sound, Upper Room Theatre Ministry; 02/2009 - 08/2011
  • I served as the sound designer for all on stage effects and engineered audio production systems for one act plays, variety shows and Broadway Style musicals.
  • Began as a stage and lighting electrician. Grew through the ranks into the Sound Designer role where I managed all aspects of audio for every major company production.
  • Audio systems were large format, totalling in 40+ channels of wireless, orchestra, and boundary microphones and SFX.
  • Negotiated with vendors to provide equipment on a strictly low budget.
  • Member of technical leadership team involved in rigging all lighting, audio and stage.
Front of House Mixer, Cox Farms; 2010
  • Mixed various Bluegrass, Rock and Jazz performances in a live outdoor style venue. 
  • Maintained electrical and audio systems of the venue. 
Sound Design Intern, George Mason University; 2009 - 2010 
  • Maintained and implemented live small and large format audio systems for sound designers 
Sound Designer, Director, and Front of House Mixer; Various musicals; 2008 – 2013.
  • Musicals: “Seussical”, “Guys and Dolls”, “Annie, the Musical”, and "Putnam County Spelling Bee".
  • Budgeted and designed a full production sound system.
  • Large format venue sound mixing.
Sound Designer; "The Beat & Path: Walk of Shame" Directed by Brandon Russell; Online Miniseries
  • Recognition: Official Selections: Miami Webfest, Toronto Webfest, Roma Webfest, NYC Webfest, DC Webfest, Indie Series Awards
  • Official site
  • Provided a full suite of services from Location Sound Mixing, and Post Production Sound Editing, Design, and Mixing.
Sound Designer; "Courting Lewis Baker" Directed by Richard Fields; Drama  
  • Recognition: Official Selections in Pan African Film Festival, Las Vegas Black Film Festival, DC Independent Film Festival, LABFF, St. Louis Black Film Festival, San Diego Black Film Festival 
  • Provided a full suite of services from Location Sound Mixing, and Post Production Sound Editing, Design, and Mixing.
Sound Designer; "Love & Variance" Directed by Richard Fields; Drama/Sci-Fi
  • Provided a full suite of services from Location Sound Mixing, and Post Production Sound Editing, Design, and Mixing.
Sound Designer; "End of the World" Directed by Sonya Dunn; Short Film
  • Provided a full suite of services from Location Sound Mixing, and Post Production Sound Editing, Design, and Mixing.
Sound Designer; “CARVE” Directed by Jonathan Jarrett; Graduate Student short film.
  • Provided Sound Design services including: Foley, Editing, ADR and Mastering.
Sound Designer; ”The Bedroom” Directed By Sonya Dunn.
  • Recognition: Official Selection Pan African Film Festival, La Femme Film Festival, and Cannes Film Festival (Short Corner).
  • Provided Sound Design services including: ADR and Mastering.
Re-Recording Mixer & Editor; “Chalk Talk” National Science Foundation; Educational Web Mini-Series.

PUBLISHED ACADEMIC WRITING

Writer & Presenter 11th Annual New Millennium Conference at American University; “The Theremin: Physics Singing Out Loud” Electronics Design/Development; Prototype Ribbon Microphone, Spring 2013
  • Built a fully functional Ribbon microphone with a hardened steel frame, neodymium magnets, and micro Aluminum foil. 
  • Developed signal processing circuit to accompany microphone. 
  • Written and public presentation of design and development.

Sound design and Re-Recording Mixing
Proudly powered by Weebly