VP of Engineering & Operations · ML Engineer · Researcher

I turn advanced research into AI systems teams can trust in the field.

I lead engineering at Analatom, where my teams build AI for defense and aerospace, from computer-vision inspection to physics-informed generative models for sonar. Researcher by training, engineer by trade, designer at heart. Based in Georgia.

See my work Download resume
Isabella Humphrey smiling, arms crossed, in front of pink azaleas

How I work

Research rigor, delivery discipline, and a good eye.

Lead engineering & operations

I run multidisciplinary teams across AI, full-stack, cloud, robotics and R&D, from the first proposal through deployment and customer handoff.

Build ML for real sensor data

Generative models, semi-supervised learning and end-to-end pipelines for acoustic, visual, depth and hyperspectral data, built for teams that rarely have enough labels.

Carry research into the field

As Principal Investigator, I take Department of Defense-funded research from early prototype to secure, deployed software people use every day.

Featured work

Inspection intelligence for aircraft that can’t afford to miss a defect.

Principal Investigator · Analatom

Automated Inspection using Deep Learning (AIDL)

A cloud platform that automates visual and subsurface inspection for defense and aerospace. AIDL pairs deep learning with hyperspectral imaging to find defects in composites faster and more accurately than traditional methods. It’s deployed on C-130 and F-15 components and moving into secure DoD cloud environments for wider MRO use.

  • Deep learning
  • Hyperspectral / SWIR
  • Cloud deployment
  • MRO
Read the case study →
SWIR inspection images beside model heatmaps highlighting defects
SWIR captures (left) → AIDL defect heatmaps (right)

Principal Investigator · Navy SBIR N251-039 Phase I

Physics-informed diffusion for sonar signal generation

A diffusion modeling framework for one-dimensional sonar signals that builds acoustic parameters from ray-based propagation models directly into the training objective. It improved alignment with expected direct-path arrival times and reduced variance in physically relevant features, without destabilizing amplitudes or degrading signal quality.

  • Diffusion models
  • Physics-informed loss
  • Mid-frequency acoustics
Cell microscopy frame beside its cilia segmentation mask

Quinn Research Group, UGA · 2019–2023

Reading ciliary motion from video

A modular generative pipeline that analyzes frame-to-frame ciliary motion to support ciliopathy diagnosis without annotated data, plus weakly supervised segmentation of cilia in biomedical video.

Pipeline: RealSense capture, fish bounding box, then depth map

UC San Diego · Semmens Lab · 2022

Counting and measuring fish underwater

An underwater RGB-D camera system, calibrated in the field, that uses AI to count fish and estimate their length for the Coastal California Fishery Research Program.

Experience

From the lab bench to the leadership table.

Full resume (PDF)
  1. Jul 2026 – now

    VP of Engineering & Operations

    Analatom Inc.

    Set engineering and operational strategy; lead AI, software and cloud development from proposal through deployment for DoD modernization programs.

  2. Feb 2024 – Jul 2026

    Principal Investigator & Technical Project Manager

    Analatom Inc.

    PI for federally funded R&D across USAF, OSD and NAVSEA programs, including the OSD-funded AIDL effort and Navy SBIR N251-039. Managed execution, reporting and deliverables as the technical interface to Government program managers, and wrote SBIR, STTR, NSF and DoD proposals.

  3. Jun – Aug 2022

    Visiting NSF Researcher

    University of California, San Diego

    Built and calibrated an AI-enabled underwater RGB-D camera for fishery research with the Semmens Lab; second author on the resulting ACM WUWNet paper.

  4. Oct 2019 – Jan 2023

    Undergraduate Research Assistant

    Quinn Research Group, University of Georgia

    Generative and semi-supervised learning for biomedical video; mentored and trained new group members.

Education

M.S. Computer Science, Georgia Tech Robotics & Perception · expected Dec 2027 B.S. Computer Science, University of Georgia AI & Data Science concentration · Aug 2023

Expertise

Generative modeling Diffusion models, VAEs, physics-informed losses Learning from limited labels Semi-, weakly supervised and active learning ML infrastructure Python, AWS, secure cloud microservices

Certifications

CITI Social & Behavioral Research

Publications

Peer-reviewed research.

  1. SciPy 2022 Incorporating Task-Agnostic Information in Task-Based Active Learning Using a Variational Autoencoder
  2. SciPy 2022 Variational Autoencoders for Semi-Supervised Deep Metric Learning
  3. ACM WUWNet 2022 Underwater Depth Calibration Using a Commercial Depth Camera (second author)
  4. SciPy 2020 Unsupervised Representation of Cilia Video Using a Modular Generative Pipeline
All publications on ResearchGate →