Sashank K.

About Sashank K.

AI/ML Engineer with around 6 years of experience building large-scale data platforms, machine learning systems, and generative-AI solutions across semiconductor, healthcare, BFSI, and engineering domains. Skilled in Python, SQL, PySpark, Kafka, TensorFlow, PyTorch, and modern cloud ecosystems (AWS, Azure, GCP). Proven track record designing real-time data pipelines, developing predictive models, and deploying production-grade AI systems using MLflow, Kubeflow, SageMaker, and Databricks. Experienced in RAG, LLM fine-tuning, vector databases (FAISS, Pinecone), and multimodal AI for computer vision and sensor analytics. Adept at collaborating with cross-functional engineering teams, improving system reliability, and delivering measurable business impact—such as reducing downtime, accelerating model deployment, and enabling rapid decision-making through automated analytics. Passionate about scalable AI, MLOps, and building enterprise-ready machine learning products.

Experience

1. Ai Ml Engineer

   Applied Materials

   May 2025 — Present · Santa Clara, CA, US

   Developed physics-informed deep learning models using TensorFlow for AIx™ platform, processing real-time sensor streams via PySpark and Kafka, improving process window stability by 60% in GAA and HBM fabrication- Built and deployed computer vision pipelines using OpenCV, CLIP, and custom CNNs on AWS SageMaker and Kubernetes, enabling automated defect classification in eBeam review tools and reducing manual review time by 45%- Designed RAG-enhanced LLM system (LlamaIndex + LangChain + private fab data in Pinecone/FAISS) for rapid materials knowledge retrieval, cutting new recipe development cycle from 6 weeks to 4 days (approximately)- Led MLOps implementation using MLflow, Kubeflow, and Airflow on Databricks Lakehouse, achieving 99.9% model deployment success rate and full traceability for ISO-compliant AI tools in production fabs- Engineered XGBoost combined with Spark MLlib predictive maintenance models on over 3 years of chamber telemetry stored in Snowflake and Redshift, reducing unplanned downtime by 32% across 200+ etch/deposition systems- Implemented Explainable AI (SHAP/LIME) dashboards in Tableau and Jupyter for model decisions in patterning and metrology, ensuring regulatory approval for AI-driven process control in customer fabs- Fine-tuned multimodal models combining sensor time-series and SEM images, deployed via Ray Serve and FastAPI, boosting overlay accuracy in hybrid bonding by 28%.

Education

• University of Central Oklahoma

  Master of Science - MS, Business Analytics (Data Science)

• Manipal Academy of Higher Education

  Bachelor of Science - BS, Business Analytics