Sarthak Hatwar

AI/ML Engineer with 4+ years of experience building and deploying production ML systems across financial services and enterprise software. Skilled in end-to-end pipelines, Generative AI, RAG systems, and multi-agent workflows using Python, PyTorch, LangChain, and Azure OpenAI. Experienced in scaling ML infrastructure on Spark and Databricks with strong MLOps and model governance practices.

AI/ML Engineer - Morgan Stanley - USA

(2025-01)

• Architected fraud detection and credit risk scoring pipelines using Python, XGBoost, and PyTorch, reducing false positive rates by 20% and achieving sub-100ms real-time inference latency across millions of daily transactions.
• Built end-to-end RAG pipelines with LangChain, Azure OpenAI, and Hugging Face to automate extraction and intelligent search across financial documents, cutting manual review time by 60% across operations teams.
• Designed multi-agent AI systems using LangGraph to orchestrate due diligence and decision-support workflows across investment banking operations, reducing analyst turnaround by 2 days per engagement.
• Scaled feature engineering infrastructure on Apache Spark and Databricks to handle 50M+ daily transactions, supporting real-time risk analytics, drift detection, and 10+ ML models deployed via MLflow and Kubernetes.
• Established SR 11-7-compliant model governance and explainability standards across 10+ production models, enabling regulatory audit readiness in collaboration with quant analysts and compliance teams.
• Spearheaded firm-wide GenAI adoption by integrating foundation models into internal tooling across investment banking, wealth management, and operations - replacing manual workflows with intelligent, scalable AI solutions.

AI/ML Engineer - Veritas Software - India

(2021-01 - 2023-07)

• Reduced unplanned system downtime by 30% by building predictive anomaly detection models using Python and Scikit-learn to proactively surface storage failures and data corruption risks inside NetBackup environments.
• Automated IT incident triage across InfoScale operations using NLP-powered log analysis and classification pipelines, cutting mean time to resolution (MTTR) by 40% and freeing support teams from manual workflows.
• Processed petabyte-scale enterprise backup datasets using Apache Spark and AWS S3 within NetBackup environments, generating storage optimization recommendations that improved disaster recovery response times.
• Embedded ML capabilities directly into NetBackup and InfoScale product lines, covering feature extraction, model training, and deployment, lifting automated data classification accuracy by 25%.
• Hardened production ML reliability through model versioning, automated CI/CD pipelines, and real-time performance monitoring across enterprise data protection and recovery systems.
• Drove continuous model improvement through A/B testing and performance benchmarking, progressively reducing false positive rates across threat detection and data classification workflows over multiple release cycles.

Masters - Data Science - The University of Texas at Arlington

Bachelors - Computer Science - MIT Art, Design & Technology University