How I can help.
What I take on, how the work runs, and what you get at the end.
01
AI product delivery
From an idea to a model running inside a product people actually use.
Most AI work dies between the notebook and the product. I take it the whole way: deciding what the model should do, getting the data, training and evaluating it, serving it, and wiring it into the app users open.
Malegado is the reference: text and speech models for 590 languages, running behind a streaming API inside a mobile app.
- Scoping what AI should and should not do
- Data collection and cleaning plans
- Models served behind real APIs
- Shipped into web and mobile apps
02
Model training and deployment
Fine-tuning, speech and text models, evaluation that means something, serving.
I train models and I put them in production. That includes low-resource languages where there is no off-the-shelf model to call, and on-device models where there is no server to call at all.
Evaluation gets as much attention as training. A number without a baseline and a held-out set is not a result.
- Fine-tuning text-to-text models
- Text-to-speech and speech-to-text
- Held-out evaluation per language or segment
- Streaming inference on a latency budget
03
Technical leadership and advisory
Leading AI teams, reviewing architectures, advising companies adding AI.
I lead the AI department at Skye8. The team reports to me, and I am still hands-on in the code, which is the only way I know to review work well.
For companies adding AI, I can look at what you are planning and tell you what will work, what will not, and what it will cost to run.
- Running an AI department
- Architecture and model reviews
- Hiring and mentoring engineers
- Honest advice on what AI will not fix