Services
Flex Analytics takes on consulting in data engineering, private AI, interactive analytics and time-series machine learning. The open-source tools it maintains are what that work is built on, and generic work flows back into them wherever a client agrees to it.
Explore years of sensor data in the browser, without exporting it or downsampling by hand. Only the aggregate a chart needs reaches the browser, so zoom and brush stay instant.
Models on long, irregular sensor and wearable signals, validated the way the domain expects. Features and models are built so a domain expert can check every step.
Sensor and operational data lands raw, auditable, and ready for analysis on a medallion architecture.
Language and vision models run where the data already is. Teams that cannot send data to a third-party API, or do not want to pay per token, get it locally.
A short call and a look at the data decide whether the approach fits and what the deliverable is.
A focused prototype or the actual integration, in the open where the client agrees to it.
Code, documentation and a walkthrough. Anything generic lands in the open-source projects.
Four rules that hold across every engagement, whatever the stack underneath.
Understand the data and the problem before proposing a solution.
The simplest thing that holds. Interpretable models where they compete with deep learning; a better query layer before a new platform.
Systems built so an agent can read and drive them: state and APIs, not screenshots. FlexViz is that idea applied to dashboards.
Design for the volume that is coming. Where speed costs accuracy or generality, the trade-off gets stated rather than buried.
Document and table OCR served on the client’s own hardware, tuned for throughput, with the data never leaving their environment.
Landing, staging and business layers, with scrapers for sources without an API and centralised data-quality checks.
Owned end to end: collection, annotation, model fine-tuning and Core ML deployment to Apple devices.
Model validation against the standard-of-care approach, and the further development that followed from it.
FlexViz, plotly-resampler, tsflex, tsdownsample and argminmax are Apache-2.0 or MIT licensed and stay that way.
Paid work funds their maintenance.
A short description of your data, what you are trying to do with it, and where the current approach stops is enough to start.
Sent. You will hear back within a few working days.
Something went wrong. Please email [email protected] instead.