Machine Learning Models
- Transformer-based architectures
- Generative pre-trained models
- Diffusion & auto-regressive models
- Variational auto-encoders
- T5, BERT, and large language models
Projects
AI research & applications
Overview
Things I'm building on my own time
Consulting pays the bills, but I also keep a couple of projects running for myself, partly to stay sharp and partly because they're problems I actually wanted solved.
They're where I try new techniques before bringing them to client work, and they're a fair sample of how I think about a problem. Two of them are below.
Kerfuffle.aiFeatured project
Featured project
Kerfuffle.ai
Small AI tools for small annoyances.
Kerfuffle is where I put little machine-learning tools for everyday problems, the kind of thing that doesn't justify a startup but is genuinely handy.
Recipe Recommender
Suggests recipes from the ingredients you have, with no lengthy backstories.
Gain Genie
A holistic fitness app generating personalized workouts for your goals.
- Tools that help you do something, not replace you
- No jargon, no sign-up wall to try them
- A testbed for ideas before they reach client work
Research platform
Lifesynai
Fake medical data that behaves like the real thing.
Medical research is hard partly because patient data is, rightly, locked down. Lifesynai generates synthetic medical records and images that hold up statistically, so researchers can build and test on them without touching anyone's real chart.
Synthetic Imaging
Realistic medical images via diffusion-modeling techniques.
Paired Data
Pairing generated images with expert-level chart-data descriptions.
- Detect medical anomalies faster and more accurately
- Generate statistically valid synthetic health data
- Maintain HIPAA compliance while enabling medical AI research
LifesynaiResearch platform
Under the hood
What these are built with
For the curious, the tools behind the projects above.
Development Stack
- Python for ML pipelines
- PyTorch and TensorFlow
- NVIDIA CUDA for GPU acceleration
- AWS for cloud infrastructure
- Modern web frameworks for interfaces
Data Management
- Secure data-processing pipelines
- Statistical validation techniques
- HIPAA-compliant data handling
- Synthetic-data quality assessment
- Data-augmentation strategies
What's next
Where I'd like to take them
Kerfuffle.ai
A few more small tools I'd like to add:
- Local business recommendation engine
- Community resource matchmaker
- Small-business workflow automations
- Personalized financial planning tools
Lifesynai
Where the synthetic-data research is headed:
- Early cancer-detection algorithms using synthetic data
- Multi-modal medical-data synthesis
- Improved statistical validation
- Healthcare workflow optimization
Work with me
Want to build something together?
If you're a business, a researcher, or a local organization with a problem that looks like one of these, I'm always up for a conversation. Some of my favorite work has started exactly that way.
Start a conversationBusinesses
A real problem you'd like to point AI at.
Researchers
Joint work with local institutions and labs.
Community groups
Nonprofits and local orgs with a tech idea.
Pilots
A small experiment to see if something's worth it.
Get in touch
Got a project in mind?
Tell me about it. Even if it's half-formed, a short conversation usually makes it clearer whether it's worth building.