Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Hi, I’m Mohammad Ramezanali, an AI/ML expert with a PhD in Theoretical Physics and over a decade of experience building and deploying AI products. I specialize in NLP, large language models, multimodal AI, and scalable cloud-based architectures (AWS, GCP, Azure). I’ve led teams delivering cutting-edge models and robust AI workflows in production environments. I've worked with industry leaders like Salesforce and Vals AI to develop AI capabilities across cloud platforms, integrate RLHF and multi-agent systems, and build end-to-end solutions that improve customer interactions and business outcomes.…Hi, I’m Mohammad Ramezanali, an AI/ML expert with a PhD in Theoretical Physics and over a decade of experience building and deploying AI products. I specialize in NLP, large language models, multimodal AI, and scalable cloud-based architectures (AWS, GCP, Azure). I’ve led teams delivering cutting-edge models and robust AI workflows in production environments. I've worked with industry leaders like Salesforce and Vals AI to develop AI capabilities across cloud platforms, integrate RLHF and multi-agent systems, and build end-to-end solutions that improve customer interactions and business outcomes.

Mohammad Ramezanali

AI Engineer, Data Scientist, Full Stack Developer, +5





Hi, I’m Mohammad Ramezanali, an AI/ML expert with a PhD in Theoretical Physics and over a decade of experience building and deploying AI products. I specialize in NLP, large language models, multimodal AI, and scalable cloud-based architectures (AWS, GCP, Azure). I’ve led teams delivering cutting-edge models and robust AI workflows in production environments. I've worked with industry leaders like Salesforce and Vals AI to develop AI capabilities across cloud platforms, integrate RLHF and multi-agent systems, and build end-to-end solutions that improve customer interactions and business outcomes.…Hi, I’m Mohammad Ramezanali, an AI/ML expert with a PhD in Theoretical Physics and over a decade of experience building and deploying AI products. I specialize in NLP, large language models, multimodal AI, and scalable cloud-based architectures (AWS, GCP, Azure). I’ve led teams delivering cutting-edge models and robust AI workflows in production environments. I've worked with industry leaders like Salesforce and Vals AI to develop AI capabilities across cloud platforms, integrate RLHF and multi-agent systems, and build end-to-end solutions that improve customer interactions and business outcomes.

Available to hire

Hi, I’m Mohammad Ramezanali, an AI/ML expert with a PhD in Theoretical Physics and over a decade of experience building and deploying AI products. I specialize in NLP, large language models, multimodal AI, and scalable cloud-based architectures (AWS, GCP, Azure). I’ve led teams delivering cutting-edge models and robust AI workflows in production environments.

I’ve worked with industry leaders like Salesforce and Vals AI to develop AI capabilities across cloud platforms, integrate RLHF and multi-agent systems, and build end-to-end solutions that improve customer interactions and business outcomes.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Language

English

Fluent

Afar

Advanced

Bashkir

Advanced

Work Experience

Lead AI Engineer at Salesforce

January 31, 2025 - January 31, 2025

Designed and developed end-to-end AI systems, NLP-based pipelines, and integrated chatbots with Salesforce CRM. Built RAG-powered retrieval and conversational AI features; led cross-functional teams to improve customer interactions and operational efficiency.

Principal AI Engineer at Salesforce

January 1, 2025 - November 12, 2025

Led end-to-end AI capability development, integrating RLHF and multi-agent systems into core product offerings, shaping AI roadmaps and presenting to C-suite stakeholders. Built RAG-enabled chatbots and context-aware assistants to improve customer interactions and deployed them into production within CRM and support workflows. Collaborated with data scientists, ML engineers, and product teams to align technical direction with business goals and deliver measurable improvements in user satisfaction.

Staff AI Engineer | Head Of AI at Vals AI

January 1, 2025 - January 1, 2025

Directed AI initiatives at Vals AI, focusing on NLP, large language models, and multi-agent systems; developed RAG-based architectures to improve retrieval accuracy and contextual responses. Integrated large language models with retrieval-based components to enhance conversational capabilities, and built production-grade models with end-to-end data processing, evaluation, and reward modeling using Docker, Kubernetes, and cloud resources on AWS and GCP. Led cross-functional teams to deliver AI-powered product features for high-profile clients, and delivered chatbots tightly integrated with Salesforce CRM to support customer service automation and improve engagement.

Head of AI at Vals AI

January 1, 2025 - January 1, 2025

Led the design and development of evaluation platforms for LLM reasoning, tool use, and agentic behavior, adopted by Anthropic, Meta, and Harvey AI, processing 50K+ model runs daily. Directed LLM fine-tuning workflows using PyTorch, Hugging Face Transformers, LangChain, and RLHF, optimizing model reasoning accuracy by 28%. Implemented cloud-based pipelines for data generation, preprocessing, and distributed evaluation on AWS, GCP, and Azure, achieving 5x throughput with 40% cost reduction. Built multi-agent simulation environments that benchmarked collaborative problem-solving, tool execution, and reasoning, producing actionable evaluation metrics for enterprise models. Developed automated dashboards and reporting systems using Python, Pandas, Plotly, and Streamlit for internal and partner-facing model analytics. Optimized LLM inference pipelines with TensorRT and Triton Inference Server, reducing latency by 35% while maintaining model fidelity. Established end-to-end CI/CD pipelines f

Principal AI Engineer at Salesforce

January 1, 2025 - November 5, 2025

Architected and implemented multi-agent LLM reasoning systems for Agentforce, enabling autonomous decision-making, tool use, and multi-step task execution, with millions of interactions processed monthly. Designed and deployed RLHF pipelines using PyTorch, JAX, and Hugging Face Transformers, achieving 3-5% improvement in agent task success rates. Built cloud-native, distributed inference infrastructure on AWS, GCP, and Azure, reducing latency by 40% for real-time agent deployment. Developed automated reward modeling and evaluation frameworks, integrated with enterprise workflows via REST APIs, gRPC endpoints, and serverless functions, and implemented scalable training and fine-tuning workflows using Ray Distributed, Kubernetes, and MLflow. Implemented monitoring dashboards with Plotly and Streamlit to visualize agent performance, reward evolution, and multi-agent interactions in real-time. Mentored cross-functional AI teams in LLM architecture, prompt engineering, multi-agent systems,

Senior Data Scientist / System Engineer / Data Engineer at CubeSmart

February 28, 2024 - February 28, 2024

Developed cloud-based data solutions, ML models, and data pipelines; performed data generation, evaluation, and reward modeling for real-time analytics; optimized data systems deployment and scalability.

Principal AI Engineer at Salesforce

February 1, 2024 - February 1, 2024

Led the design and implementation of scalable multi-agent AI systems that leverage NLP techniques, LLMs, and retrieval-augmented generation (RAG) to autonomously perform complex tasks across diverse customer interactions, seamlessly integrated into Salesforce's cloud environment. Architected AI-driven workflows by optimizing the agent decision-making process with reinforcement learning, improving conversation quality and reducing manual intervention. Built and deployed production-grade AI pipelines—from data generation and evaluation to model tuning and continuous deployment—using PyTorch, TensorFlow, FastAPI, and cloud runtimes on AWS, GCP, and Azure. Drove real-time inference capabilities with Docker and Kubernetes and helped define CI/CD pipelines, ensuring reliable, scalable delivery. Collaborated with product and partner teams to align AI capabilities with Salesforce's business goals and customer outcomes.

Lead AI Engineer at Salesforce

January 1, 2024 - January 1, 2024

Designed and deployed the Technical Health Score (THS) pipeline, tracking $1B+ in ML-powered deployments, providing real-time observability and decision support for enterprise-scale AI systems. Developed and maintained internal ML platform for hundreds of NLP and LLM models, enabling standardized training, evaluation, and deployment across teams. Led NLP initiatives including sentiment analysis, intent recognition, entity extraction, summarization, and question-answering using PyTorch, Hugging Face Transformers, spaCy, and NLTK, achieving 20–35% accuracy gains over baseline models. Implemented cloud-based training and inference pipelines on AWS, GCP, and Azure, processing terabytes of data with distributed computing frameworks. Architected embedding-based search, recommendation, and vector retrieval pipelines using FAISS, Pinecone, and ElasticSearch, reducing query latency by 60%. Deployed scalable inference pipelines with Docker, Kubernetes, MLflow, and Triton Inference Server, achi

Senior Data Scientist at CubeSmart

December 1, 2018 - December 1, 2018

Developed predictive pricing, customer churn, and demand forecasting models using Python, TensorFlow, PyTorch, and scikit-learn, driving 15% revenue growth year-over-year. Built and deployed LSTM and Transformer-based NLP models for text classification, recommendation, and personalization, improving customer targeting accuracy by 22%. Implemented cloud-based ETL pipelines on Azure and AWS, automating data preprocessing and feature engineering workflows, cutting manual workload by 70%. Designed vector-based recommendation engines leveraging embeddings and similarity search, increasing acquisition efficiency by 25%. Created executive dashboards using Python, Pandas, Plotly, and Streamlit for actionable business insights, predictive KPIs, and model interpretability. Optimized model inference pipelines using Docker, Kubernetes, and serverless architectures, enabling real-time predictions for customer-facing applications. Introduced synthetic data augmentation techniques to enrich training