CUDA and Go GPU processing for an AI company
Moving heavy data processing onto NVIDIA GPUs with CUDA, driven from Go, for an AI business working with large datasets.
The brief
The client, who runs an AI business, needed processing moved onto NVIDIA GPUs using CUDA, controlled from the Go language, for workloads expected to involve large datasets. Their own brief described it as a challenging, puzzle-like project, and few developers work in both CUDA and Go.
What we built
- GPU processing code written in CUDA
- Integration between the Go application and the GPU code
- Parallel processing design for large data volumes
How it was delivered
Scoped and delivered as a fixed piece of work, with the client testing on their own hardware.
Result
The client rated it five stars and wrote that not a lot of people can do this work.
This is the layer under every serious AI deployment: making models and data processing run fast on the hardware you already pay for.
Voxlivo: the AI receptionist we built and run
An AI phone receptionist for plumbing and HVAC companies. It answers in about two seconds, day and night, and books the job.
Modular data acquisition platform in LabVIEW
One application that acquires, displays, logs and alarms on data from test hardware, built to grow as new instruments are added.