Ali S.

About Ali S.

Built and led production AI systems where reliability, governance, and cost discipline…

Experience

1. Principal Ai Ml Architect - Llms & Reasoning Systems

   IBM

   Mar 2022 — Present

   Architected and deployed large-scale LLM and reasoning model pipelines using PyTorch, TensorFlow, HuggingFace Transformers, and process reward models (PRMs) to support advanced compliance and domain-specific reasoning tasks in production.• Implemented Retrieval-Augmented Generation (RAG) systems combined with Knowledge Graph (KG) reasoning techniques (A2N-style), enabling context-rich and fact-aware enterprise search in regulated industries such as finance and healthcare.• Led distributed training and optimization of Mixture-of-Experts (MoE) models exceeding 100B parameters, leveraging AWS p4d instances, FSDP, and ZeRO Stage 3 for efficient scaling and stable convergence in HPC cloud environments.• Designed and operationalized meta-learning task generators from unlabeled enterprise datasets, enabling few-shot adaptation across multiple NLP tasks and reducing turnaround time for deploying specialized models.• Directed an AI/ML engineering and research team responsible for rapid prototyping, building evaluation harnesses (e.g, BBH, MATH, GPQA), and iteratively aligning models to domain requirements through structured experimentation.• Partnered with product, platform, and data engineering teams to integrate multi-agent RL simulations for policy optimization and conversational strategy development, extending model capabilities for interactive systems.• Established secure, scalable AI infrastructure leveraging AWS CloudFormation, Azure IAM, and container orchestration platforms, aligning with SOC 2, HIPAA, and GDPR compliance requirements.• Collaborated closely with business stakeholders during pre-sales and delivery phases to define AI/ML architectures, produce technical solution blueprints, and ensure successful integration into client ecosystems.• Provided ongoing advisory, best practices, and quality governance for AI/ML initiatives to ensure operational reliability, maintainability, and alignment with evolving compliance standards.