Goutham Katta

Ai Ml Engineer @Scale AI

US

Worked on Large Language Model (LLM) development and evaluation pipelines, supporting enterprise and autonomous-system customers by improving training data quality, model alignment, and performance benchmarking- Designed and implemented human-in-the-loop data pipelines for LLM training, including data labeling, quality control, and feedback loops to support supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF)- Built and optimized Retrieval-Augmented Generation (RAG) workflows by integrating embedding models with vector databases, enabling grounded and context-aware responses for knowledge-intensive GenAI applications- Applied NLP techniques such as text normalization, embedding generation, prompt structuring, and sequence modeling to enhance downstream LLM training, evaluation, and inference workflows- Contributed to Generative AI evaluation frameworks, benchmarking open-source and proprietary LLMs across dimensions such as accuracy, hallucination rate, safety, and response quality- Supported multimodal AI systems involving text, image, and video data, enabling data pipelines and evaluation workflows for autonomous vehicles, simulation, and synthetic data generation use cases- Developed scalable data processing pipelines using Python, Spark, and Airflow to manage large-scale structured and unstructured datasets used in LLM training and model validation- Collaborated with platform and infrastructure teams to productionize ML workflows using Docker, Kubernetes, and MLflow, ensuring reproducibility, experiment tracking, and scalable deployments- Worked with cloud-based GPU infrastructure (AWS) to support distributed training, evaluation, and CI-driven testing of LLMs and multimodal models- Partnered with cross-functional teams including research, product, and operations to translate customer requirements into high-quality training data and robust ML solutions.