Deepak K

About

ML/LLM engineer with production experience building a compliant RAG-based virtual assistant at Intuit, optimizing embeddings and FAISS retrieval (including PCA) for low-latency, privacy-controlled search and deploying via AWS SageMaker containers. Also built scalable Airflow+MLflow pipelines using Docker and KubernetesExecutor, cutting training cycles by 37%, and partnered with civil engineers/project managers at Aegis Infra to deliver predictive maintenance for construction equipment.

Key Strengths

• Built and deployed a production RAG virtual assistant for financial customer support with compliance and privacy considerations
• Optimized retrieval performance via embedding model selection, PCA dimensionality reduction, and FAISS vector search
• Designed scalable ML orchestration with Airflow + MLflow; resolved dependency/environment issues using Docker and KubernetesExecutor
• Reduced model training cycles by 37% through pipeline reliability and scalability improvements
• Strong evaluation discipline: KPI-driven design, A/B testing, human-in-the-loop review, and real-time monitoring with Prometheus
• Effective cross-functional collaboration with non-technical stakeholders using dashboards/visuals and weekly syncs

Experience

Education

Skills

Languages