Vineeth Sanke

About Vineeth Sanke

I’m a AI/ML Engineer and Machine Learning Engineer specializing in Generative AI, NLP, and scalable data systems, with 5+ years of experience building production-grade ML and GenAI solutions across healthcare and financial services. My work focuses on designing and deploying end-to-end AI systems from data ingestion and feature engineering to model training, fine-tuning, evaluation, and large-scale deployment. I have hands-on experience building LLM-powered applications, including Retrieval-Augmented Generation (RAG) pipelines, transformer-based architectures, and multi-agent AI systems for real-world use cases. At Spring Health, I developed scalable ML and GenAI pipelines processing millions of records to support clinical and financial reporting. I built robust RAG-based architectures with strong guardrails to improve response accuracy, reduce hallucinations, and ensure reliability in sensitive healthcare environments. Core Expertise: • Generative AI & LLMs (OpenAI, Hugging Face, Ollama, Amazon Bedrock, NVIDIA NIM) • Fine-tuning LLMs (LoRA, QLoRA) • RAG Pipelines, LangChain, Embeddings, Hybrid Search • Multi-Agent Systems (CrewAI, LangGraph) • NLP & Deep Learning (Transformers, LSTM, GRU, ANN) • Machine Learning & Predictive Modeling • Data Engineering (PySpark, Airflow, Snowflake) • Cloud Platforms (AWS, Azure) I’m passionate about building AI systems that are not only powerful, but also scalable, reliable, and production-ready.

Experience

1. Ai Ml Engineer

   Spring Health

   Sep 2024 — Present · NY, US

   Scalable LLM and RAG Infrastructure: Designed and implemented high-throughput data ingestion pipelines using AWS Glue and Amazon SageMaker, processing 5M+ records into optimized Parquet formats to power Retrieval-Augmented Generation (RAG) systems, reducing LLM inference latency by 50%. • LLM Data Pipeline and Embedding Engineering: Built end-to-end NLP preprocessing and embedding pipelines, transforming large-scale unstructured clinical text into tokenized, vectorized, and embedding-ready datasets, reducing data preparation time by 40% and enabling efficient semantic retrieval. • Production-Grade AI API Development: Developed scalable RESTful APIs using FastAPI to serve LLM and SageMaker endpoints, implementing semantic caching and optimized retrieval strategies to improve response time by 35% and reduce compute costs. • Generative AI MLOps and Orchestration: Architected automated CI/CD pipelines for LLM workflows using SageMaker Pipelines and AWS Step Functions, ensuring reproducibility of prompts, embeddings, datasets, and model outputs across environments. • Vector Database and Feature Store Integration: Designed and maintained a centralized feature store and embedding repository (SageMaker Feature Store), enabling efficient storage, retrieval, and reuse of vector embeddings for RAG-based applications. • AI Observability, Guardrails and Monitoring: Implemented LLM monitoring and governance using Python and SageMaker to detect data drift, schema anomalies, bias, and hallucinations, integrating webhook-based alerting systems for real-time model reliability in production environments.

Education

• Pace University

  Master's degree, Computer Science