taking projects · checking…
Gradient Works
Applied AI research and model engineering.
An applied AI studio I founded. It designs, trains and tests custom machine learning models for startups, research labs and small teams: every engagement has a defined metric, a baseline to beat, held-out testing and a written result, and the client keeps all the code and weights.
- the problem
- Startups, labs and small teams need custom models but rarely have the research depth to build them and evaluate them honestly.
- what I built
- A studio with a fixed-scope, four-week process: a scoping call, a baseline to beat, tracked experiments, held-out testing and a full handover of code and weights.
What it does
- Custom model builds: tabular prediction, computer vision, time series and biosignals, small LLM fine-tuning
- Testing and evaluation: benchmark design, bias and robustness checks, LLM evals and red-teaming, audit reports
- Research and prototyping: literature reviews, paper reproduction, baselines and feasibility studies
- AI-native features: RAG over your documents, agents and automation, model APIs, data pipelines
- A four-week sprint from a free scoping call to handover, with tracked experiments shared every week
- Fixed scope and price; the client owns the code, weights, notebooks and report, with nothing hosted on our side
- A lab that ships open models between client projects, starting with NeuroDial for EEG classification
How it's built
each layer feeds the one below
Scope
- free 30-minute call
- problem statement and success metric
- data audit
Build
- baseline model and reference score
- tracked experiments, weekly updates
Ship
- held-out testing and failure analysis
- code, weights and evaluation report
- handover call
PyTorchHugging FaceFastAPI