Research
All topics
Every topic under which Dingal AI Research publishes, with the number of entries in each.
All research topics
Inference engines
Building model runtimes around the constraints of hardware people can own and operate.
4 entriesMemory hierarchy
Treating graphics memory, system memory and fast storage as one managed resource.
3 entriesQuantisation
Reducing the weight of expert storage while measuring the quality that is traded away.
2 entriesSpeculative decoding
Testing faster decoding paths that must preserve the output of the verified path.
1 entryResearch practice
Pre-registered thresholds, correctness controls, negative results and decisions that survive the result.
4 entries