Rahul Hatkar

Designed and deployed Python-based enterprise AI agent workflows leveraging MLflow, Mosaic AI Agent Framework, Vector Search, and Delta Lake. Led RAG optimization, embedding tuning, and automated evaluation pipelines, boosting response accuracy by 22% and cutting inference cost by 31%. Reduced end-to-end latency by 38% through retrieval optimization, vector indexing improvements, asynchronous processing, and scalable serving pipelines. Built agent evaluation frameworks (LLM-as-a-Judge) for groundedness and tool accuracy. Engineered scalable retrieval-augmented generation pipelines and cloud-native microservices on AWS using Docker, Kubernetes, and Databricks Model Serving. Implemented automated MLOps pipelines with MLflow, Delta Lake, and CI/CD, enabling governance and high-scale operations.

Developed Python-based distributed evaluation pipelines processing over 2 million benchmark runs monthly, improving model accuracy by 18% on coding and reasoning tasks. Built automated safety and agentic evaluation frameworks, reducing validation cycles by 42% and supporting frontier AI deployment. Created synthetic data generation and quality-filtering workflows increasing benchmark coverage by 35%. Built ML pipelines with PyTorch, JAX, and Ray for large-scale experimentation; implemented RL and preference optimization workflows. Developed end-to-end evaluation services with Python, REST APIs, PostgreSQL, and containerized workloads. Architected microservices-based evaluation platforms on Kubernetes, AWS EKS, and cloud-native data processing on S3/Lambda/Batch. Implemented observability via Prometheus, Grafana, and OpenTelemetry, reducing incidents and improving reliability. Optimized serving infrastructure to support 500k+ monthly users with cost reductions.

Delivered end-to-end ML pipelines using Python, PySpark, Scikit-Learn, and MLflow for 2.5M+ users, boosting prediction accuracy by 12%. Built feature stores and automated retraining to reduce latency by 28% and accelerate deployment. Implemented scalable data processing with Databricks, Spark, Kafka, Snowflake for real-time analytics across multiple domains. Designed cloud-native AI solutions on AWS (S3, EMR, Lambda, SageMaker) and Terraform; built microservices-based inference with FastAPI, Docker, Kubernetes, Redis. Automated CI/CD and MLOps using AWS, GitHub Actions, Jenkins, Airflow, MLflow Registry and CloudWatch. Developed enterprise data lake/lakehouse with Delta Lake and Spark to support batch and streaming analytics.