Muppa Sai Vinay

Analytical AI Training Specialist with proven expertise at RWS Group and Innodata Inc. delivering high-fidelity datasets for top Tier-1 technology clients across both companies. Experienced in optimizing Large Language Models (LLMs) through Reinforcement Learning from Human Feedback (RLHF), pairwise ranking, and chatbot fine-tuning. Versatile across multimodal domains, including Speech (TTS/STT) and Computer Vision (OCR) for Telugu and English, to maximize AI model safety, accuracy, and performance.

Data Annotator - Telugu at RWS Group (2026-04 – Present)

• Project 1: Generative AI Chatbot Training - Conversational AI Alignment: Trained and fine-tuned multilingual AI conversational models across specialized domains, including e-commerce, healthcare, and customer service scenarios to ensure natural, domain-specific user interactions.
• Project 2: Computer Vision and Telugu OCR (Document Digitization) - Optical Character Recognition (OCR): Executed high-precision image-to-text transcription and bounding-box annotation for complex Telugu script found in scanned documents and real-world images.
• Project 2: Computer Vision and Telugu OCR (Document Digitization) - AI Training Datasets: Curated and labeled specialized training datasets to optimize OCR model performance across healthcare, education, and public administration documentation.
• Project 3: Multimodal Speech, Text Evaluation and LLM Ranking - Data Annotation and Ranking: Conducted critical pairwise comparisons and response ranking tasks to optimize LLM performance, safety, and helpfulness through reinforcement learning feedback loops.
• Project 3: Multimodal Speech, Text Evaluation and LLM Ranking - Linguistic Quality Assurance: Evaluated generative AI text for grammatical precision, contextual accuracy, and cultural appropriateness while performing rigorous bidirectional translation quality assessments for English and Telugu.
• Project 3: Multimodal Speech, Text Evaluation and LLM Ranking - Text-to-Speech (TTS) Optimization: Evaluated synthetic Telugu voice model outputs for nativeness, tonal naturalness, and intonation cadence to improve high-fidelity AI speech generation.
• Project 3: Multimodal Speech, Text Evaluation and LLM Ranking - Speech-to-Text (STT/ASR) Training: Transcribed diverse Telugu audio datasets into high-accuracy text to train and refine Automatic Speech Recognition (ASR).

Data Annotator at Innodata (2025-11 – 2026-03)

• Model Alignment: Conducted high-volume RLHF tasks for a global technology leader, ranking AI-generated responses based on nuance, tone, and logic.
• Adversarial Red-Teaming: Authored and tested 1,000+ adversarial prompts to identify model vulnerabilities in safety, bias, and policy compliance, maintaining a 95%+ quality threshold.
• Platform Expertise: Mastered enterprise-level multi-modal annotation tools to manage complex workflows involving text, logic, and safety filtering.
• Quality Excellence: Maintained a consistent QA score of 95%, adhering to 50+ pages of evolving project guidelines.

Bachelor Of Technology in Electronics and Computer Engineering – JB Institute Of Engineering And Technology (2019-08 – 2023-08)