Build and fine-tune LLM/ML models for Arabic NLP, document classification, vision/OCR, and AIOps use cases.
Run pre-deployment evaluation
Accuracy baselines, regression and safety testing; evidence to justify GPU allocation.
Optimize inference — quantization, batching, context sizing — against measured usage.
Deploy on Humain GPUaaS: Kubernetes, GPU partitioning on B300 nodes, quotas, RBAC.
Build equivalent workloads on GCP (Vertex AI, GKE) with classification-based routing.
Own serving stack (vLLM/TGI), model versioning, CI/CD, and monitoring for latency, tokens, GPU utilization, and drift.
Ensuring developed AI Models Complying with ZATCA data sovereignty and SDAIA requirements (AI Ethics, GenAI Guidelines, PDPL).