Product Manager and lead AI/ML Engineer for "Terror Bot" -- Breacher.ai's in-house Agentic AI Impersonator; which uses a cloned voice and carries out tailored, real-time social engineering conversations to simulate deepfake attacks for our clients. "Terror Bot" averages a ~17% rate of socially engineering users into taking actions outside of corporate policy.

I designed and built the Breacher's audio data curation and pre-processing app. The python application — “Terror Forge” — ingests OSINT audio of targets, trims silence, normalizes volume, cuts out crosstalk, and separates speakers into individual clips. The output is ready made training data for voice clones for our clients. This application reduced the time spent preparing data for voice cloning from 1-2 hours per clone to ~5 minutes.

Building an in-house voice cloning text-to-speech model and training pipeline for use in real-time voice agents. Using opensource datasets, PyTorch Lightning 2x, NeMo models, torchaudio, OmegaConf, librosa.

Develop strategies for enterprise clients to improve cognitive defense and cybersecurity posture against AI-powered social engineering attacks.

Create Red Team Simulation campaigns for clients.

Terror Forge – Voice AI training data prep pipeline (Python, pyannote.audio, diarization). Automates speaker-isolated audio extraction from YouTube. GitHub

Terror Root – Built a full TTS training pipeline using NVIDIA NeMo FastPitch. Implemented data prep (audio/text alignment, pitch extraction, tokenizer setup, 7-tuple formatting), enabling successful training and model loading.

Deepfake Artist in Residence

Reality Defender detects deepfakes and AI-generated content for enterprises, platforms, and governments.

Silvereye is a Cloud Distributor with a division that does AI strategy and systems integration, another that does cybersecurity value-added reselling and marketing, for the US Government, Federal Systems Integrators, the Defense Industrial Base and the Intelligence Community.

Sourced and closed a $1M+ cybersecurity deal with the NNSA (DOE) for OEM partner.

Led ETL project that collated all visually confirmed Russian equipment losses from the Russo-Ukraine war into a structured dataset (available on my GitHub link) using BeautifulSoup, Python, Pandas, and DataFrames.

Created viral marketing memes (see examples from my portfolio on this page) that have seen 18%+ response (sic) rates. I have created, executed on, and received follow on contracts for marketing campaigns that total ~120k per year of marketing development funds from our partners, including Carahsoft.

I lead our work to solve the automatic speech recognition problem for the FAA and USAF, and our allies. Ask me or the leadership at www.gov-acq.com for more on how to unlock air traffic speech.

Sourced and closed $1M+ in net new business with the US Government in 2023.

Onboarded and managed first inside sales rep hire; who sourced ~$150k in new pipeline in 2023, and $180k in 2024.

Founding Federal sales rep, 17th employee of an In-Q-Tel portfolio company.

As a business scientist, I curated datasets to train deep learning models to unlock critical speech data to support missions within DoD and Civilian agencies.

Sourced and closed a $1M dollar deal with the Intelligence Community.

Landed 1st Federal System Integrator deal.

Hunted 1st allied Government deal.

Through my work, Deepgram, our partners, and NASA produced the best performing speech recognition model on NASA’s space-to-ground audio. At 94%+ we become the very first vendor to score higher than 78% (WRR) Word Recognition Rate on NASAs validation set of astronaut speech data from the ISS. Thousands of NASA acronyms? Done. Landed NASA logo, then expanded to multiple NASA organizations, including Neutral Buoyancy Lab. Our work garnered a panel and speaking spot at NVIDIA GTC 2022.

Partner manager for; In-Q-Tel, NVIDIA, Carahsoft, Government Acquisitions Inc.

Registered Deepgram on Sam.gov and for a CAGE code.