# Engineering - From your workflow down to the GPU kernel

> How we design, adapt, serve and tune the systems we build: systems architecture, model adaptation and evaluation, inference performance and GPU kernels.

- Web page: https://www.alesanalytics.com/engineering
- Markdown: https://www.alesanalytics.com/engineering.md
- Publisher: Ales Analytics

## From your workflow down to the GPU kernel.

An AI project can stop at plugging a hosted model into a workflow. We work further down as well: the workflow and applications, the data, the models, the serving stack and the hardware they run on. Each layer is designed and tested for the job it has to do.

## The layers we work across

1. **Workflows and applications.** The process as your team runs it: the approvals, the systems it touches, the screen people use.
2. **Data and retrieval.** Pipelines, canonical records, document reading, and search that respects permissions.
3. **Models.** Open-weight or hosted, chosen and evaluated on examples from your own work.
4. **Model adaptation.** Fine-tuning, preference tuning and distillation, where evaluation shows a gap worth closing.
5. **Serving and inference.** Engines, batching, caching, quantisation and capacity sized for your GPUs.
6. **GPU kernels and hardware.** Profiling first, then fused and attention kernels where a bottleneck shows up.

## Read the engineering notes

- [Principles](https://www.alesanalytics.com/engineering/principles.md): The nine rules every build follows
- [Systems architecture](https://www.alesanalytics.com/engineering/systems-architecture.md): Documents, retrieval, workflows, data and deployment, as one design
- [Model adaptation and evaluation](https://www.alesanalytics.com/engineering/model-adaptation.md): Choosing, adapting and measuring models on your own data
- [Inference performance](https://www.alesanalytics.com/engineering/inference.md): Serving engines, scheduling, quantisation and capacity sizing
- [GPU kernel engineering](https://www.alesanalytics.com/engineering/gpu-kernels.md): Rooflines, CUDA, Triton, attention kernels and verification

## Get ahead with Ales Analytics

Tell us what you want to build. Our engineers take it from the first conversation to a system your team relies on every day. https://www.alesanalytics.com/contact

## Contact

solutions@alesanalytics.com, or https://www.alesanalytics.com/contact.

- Every page as Markdown: https://www.alesanalytics.com/llms.txt