AI & ML interests
Speech technologies, NLP, CV, Machine Learning, Deep Learning, Reinforcement Learning, AI for medicine, AI computational resources, AI in robotics, AR/ VR in AI
Recent Activity
A 0.6B edge LLM that emits calibrated verbalized confidence for edge-cloud routing. Models + training/eval data.
Collection of FontID related repos
These are the vision benchmarks used to evaluate Qolda model
Kazakh-adapted Qwen3.5 (4B / 9B / 35B-A3B): instruct models and their continued-pretrained base on extended vocabulary
Bilingual Kazakh–English target-speaker ASR for overlapping speech — checkpoints plus the KazMix-3 and PersonaMix datasets. The first Kazakh TS-ASR.
These are the audio benchmarks used to evaluate Qolda-AVL model
These are the language benchmarks used to evaluate Qolda model
Kazakh-adapted Qwen3.5 (4B / 9B / 35B-A3B): instruct models and their continued-pretrained base on extended vocabulary
Bilingual Kazakh–English target-speaker ASR for overlapping speech — checkpoints plus the KazMix-3 and PersonaMix datasets. The first Kazakh TS-ASR.
A 0.6B edge LLM that emits calibrated verbalized confidence for edge-cloud routing. Models + training/eval data.
These are the audio benchmarks used to evaluate Qolda-AVL model
Collection of FontID related repos
These are the language benchmarks used to evaluate Qolda model
These are the vision benchmarks used to evaluate Qolda model