Sabarish Ravichandran Mehala

Improved demand forecasting accuracy by ~8% MAE by designing and optimizing time-series ML models with advanced feature engineering, translating business needs into scalable predictive solutions for operational planning. Built and deployed end-to-end ML pipelines on AWS (S3, InfluxDB), integrating monitoring and KPI tracking via Grafana to ensure measurable business impact and continuous performance evaluation. Developed robust, production-ready data pipelines with automated data validation, reproducible workflows, and version control using Git, enabling reliable deployment, testing, and continuous improvement aligned with enterprise data science practices.

Designed and deployed scalable data pipelines for document intelligence (OCR + LLM + RAG) on Kubernetes, translating business needs into efficient AI-driven workflows and reducing processing time by 65%. Built a modular multi-agent system using FastAPI and Pydantic with fallback orchestration and shared memory, improving system reliability from 92% to 99.5% and enabling robust, context-aware decision-making. Transformed research prototypes into production-ready microservices with a focus on scalability, observability, testing, and fault tolerance, aligning with enterprise standards for deployment, monitoring, and continuous improvement.

Developed predictive models combining time-series data with exogenous signals (e.g., weather APIs), improving forecast accuracy by 18% while applying rigorous validation techniques such as anomaly detection and resampling to ensure model reliability. Designed decision-oriented optimization pipelines linking predictions to consumption patterns, enabling ~12% CO₂ reduction and cost savings, and defined clear KPIs using a North Star framework to align AI solutions with measurable business impact.