AI/ML Engineer at Snowflake

Orchestrated enterprise agentic AI workflows using Python, Snowflake Cortex Agents, MCP, FastAPI, and governed access via Snowflake SQL, handling 250+ daily tool-calling requests with human approval and policy enforcement. Architected agentic RAG combining Cortex Search, BM25, dense embeddings, hybrid retrieval, and reranking to improve retrieval relevance by 10% for governed structured and unstructured knowledge applications. Operationalized AI evaluation and observability using Cortex Agent Evaluations, Cortex AI Observability, MLflow, and LLM-as-a-judge, shortening regression-analysis cycles by 15% through golden datasets, tracing, and quality gates. Secured production agents using Horizon Catalog, RBAC, Cortex AI Guardrails, Pydantic, and audit logging with prompt-injection protection and human approval. Deployed production AI/ML services using Snowflake, Docker, Kubernetes, OpenTelemetry, Prometheus, and Grafana, coordinating with platform and security teams to sustain 99.5% avail

Machine Learning Engineer at Freshworks

Built ticket classification, routing, and SLA-risk models with Python, SQL, scikit-learn, XGBoost, and BERT, reducing manual triage effort by 8% across 300+ daily tickets. Improved knowledge recommendation pipelines using Sentence-BERT, FAISS, Elasticsearch, and embeddings, increasing article relevance by 9% and supporting workflows for 75+ support agents. Delivered 2023 generative AI workflows using GPT-3.5-turbo, GPT-4, and Azure OpenAI Service with prompt engineering and RAG, reducing agent handling effort by 11% and contributing $18K+ in annualized efficiency savings. Productionized ML services with FastAPI, REST APIs, Docker, Git, and MLflow to support 400+ daily workflows while maintaining 92.4% availability via deployment discipline, model versioning, drift monitoring, retraining, and release validation. Collaborated with product, customer support, data engineering, and backend teams to translate requirements into ML objectives, API contracts, evaluation criteria, and feedback l