Hire Chaitanya - Fine-Tuning Engineer

* Designed and implemented rigorous AI model evaluation pipelines, improving model safety detection accuracy by * 32%, reducing potential deployment errors significantly across enterprise AI systems. * Developed scalable automated benchmark systems for evaluating reasoning and alignment, increasing evaluation * throughput by 45%, and enhancing model reliability across multiple LLM frameworks. * Integrated human-in-the-loop evaluation workflows for safety, hallucination detection, and bias assessment, * achieving 28% faster feedback cycles and actionable alignment improvements. * Engineered large-scale distributed training pipelines leveraging PyTorch, JAX, and TensorFlow, optimizing * transformer-based LLM performance with efficient GPU and mixed-precision strategies. * Applied RLHF and advanced generative AI techniques to fine-tune models, incorporating multi-agent evaluation * frameworks for robust reasoning, alignment, and instruction-following capabilities. * Designed adversarial testing frameworks to simulate edge-case scenarios, leveraging automated scoring metrics, * model-as-judge evaluations, and statistical performance analysis. * Collaborated cross-functionally with researchers, product managers, and data annotators, demonstrating structured * problem-solving, mentorship, and Agile leadership in complex AI/ML deployments. * Developed modular evaluation pipelines for NLP and computer vision tasks, ensuring reproducible, high-quality * benchmark data and seamless integration with enterprise AI platforms. * Deployed cloud-based AI evaluation systems on AWS, optimizing resource allocation, monitoring distributed * workloads, and ensuring scalable model testing across multiple environments. * Implemented secure, containerized evaluation workflows on AWS, integrating metrics automation, continuous * benchmarking, and multi-agent orchestration to improve alignment and model safety.

* Built and optimized large-scale training pipelines using PyTorch and Hugging Face, improving model accuracy by 21% through fine-tuning, * data optimization, and distributed GPU training workflows. * Developed distributed training systems using PyTorch FSDP and GPU clusters, improving memory efficiency, training stability, and batch * processing performance for transformer-based architectures. * Designed evaluation and benchmarking workflows using Python and MLflow, improving experiment tracking consistency by 34% and * accelerating model comparison across multiple datasets and training runs. * Worked on parameter-efficient fine-tuning techniques including LoRA and PEFT to adapt large language models for domain-specific tasks * while reducing overall compute and training overhead. * Built retrieval and context-enrichment pipelines to improve response relevance, information grounding, and downstream model * performance across production AI applications. * Optimized large-scale training and inference workflows by improving GPU utilization, data throughput, and distributed execution efficiency * across high-performance compute environments. * Developed scalable backend APIs and inference services using Python and FastAPI, enabling seamless integration of ML models into * distributed microservices and production systems. * Engineered streaming and batch data pipelines using Apache Spark and Kafka, increasing data throughput by 26% while supporting scalable * training, inference, and experimentation workflows. * Collaborated with ML engineers and research teams on model experimentation, training optimization, and deployment strategies for largescale AI systems and distributed infrastructure. * Deployed containerized ML applications using Docker and Kubernetes, managing orchestration, CI/CD pipelines, and scalable cloud * deployments for reliable production-grade ML operations.

* Built machine learning workflows in Python and SQL for forecasting, anomaly detection, and operational analytics across enterprise business datasets. * Improved model and reporting accuracy by 32% through data preprocessing, feature preparation, and optimization of large-scale structured datasets. * Reduced manual effort by 45% by automating ETL processes, validation workflows, and recurring data preparation tasks used in analytics and ML pipelines. * Worked with AWS services including S3 and Redshift to support scalable data ingestion, transformation, and machine learning workflow execution. * Performed exploratory data analysis, trend analysis, and statistical validation on structured and semi-structured datasets to support predictive modeling initiatives. * Developed Python-based forecasting and anomaly detection models to identify operational trends and support business planning and decision-making processes. * Collaborated with engineering, analytics, and stakeholder teams to support deployment and improvement of ML-driven reporting and automation solutions. * Improved scalability and reliability of ML and analytics workflows through automation, query optimization, and monitoring of enterprise data pipelines