Samruddhi Shirbhate

AI Research Associate and Prompt Engineer with 1-year experience in prompt engineering, prompt optimization, AI model training, and response evaluation across Large Language Model (LLM) workflows. Skilled in designing and refining prompts for Generative AI systems, conducting data annotation, dataset evaluation, and improving AI response quality through hallucination detection and RLHF-aligned feedback loops. Proficient in Python, Natural Language Processing (NLP), and Retrieval-Augmented Generation (RAG)-based workflow design, with a consistent focus on delivering accurate, reliable, and user-centric AI solutions.

Research Associate - AI - Keywords Studios - Bangalore, Karnataka, India

(2025-07)

• Designed and optimized prompts for 40+ Amazon Nova model use cases, improving response quality, accuracy, and task alignment across diverse scenarios
• Contributed to AI model training by refining and evaluating 500+ prompts, enhancing model performance and instruction-following capability
• Assessed 1,000+ AI-generated outputs weekly to identify hallucinations and factual inconsistencies, improving overall response reliability
• Developed multi-step conversational flows and prompting strategies that supported complex user interactions across 3 business scenarios
• Built test cases and evaluation frameworks adopted team-wide to validate model behavior and ensure consistency, safety, and compliance
• Executed data annotation, content review, and quality validation for large-scale AI training and evaluation pipelines
• Collaborated with product, engineering, and research teams to translate user feedback into prompt optimizations and scalable AI solutions

Bachelor of Engineering - Information Technology - Sipna College of Engineering and Technology (2020-06 - 2024-05)