Pavan Sai Merugumala

Associate Analyst – Google Client at Vaco Binary Semantics LLP

Prepared audit reports and quality statistics for human raters on AI evaluation tasks. Automated data handling in Google Sheets using Google Apps Script to improve process efficiency. Contributed to developing an AI model analyzing image-topic relevance based on scraped web data. Evaluated AI responses using Reinforcement Learning with Human Feedback (RLHF) under a Human-in-the-Loop (HITL) framework.

Freelance AI Trainer – Coding Expert at Outlier AI

Debugged and assessed LLM-generated code responses for DSA and full-stack problems (HTML, CSS, JS). Tested responses in IDEs like VS Code to identify bugs and ensure instruction-following. Designed and evaluated prompts using prompt engineering techniques to optimize LLM outputs. Created adversarial prompts to test model failure cases and evaluate factuality of generated content.

Freelance AI Trainer – Bilingual (Telugu) at Appen

Reviewed and rated AI-generated Telugu responses for linguistic accuracy, fluency, and prompt alignment. Evaluated translations based on grammar, completeness, cultural sensitivity, and context.

Associate Analyst at Vaco Binary Semantics LLP

Prepare audit reports and quality statistics for human raters on AI evaluation tasks. Automate data handling in Google Sheets using Google Apps Script, improving process efficiency. Contribute to the development of an AI model that analyzes image-topic relevance based on scraped web data. Evaluate AI responses using Reinforcement Learning with Human Feedback (RLHF) under a Human-in-the-Loop (HITL) framework.

Freelance AI Trainer at Outlier AI

Debug and assess LLM-generated code responses for DSA and full-stack problems (HTML, CSS, JS). Test responses in IDEs like VS Code to identify bugs and ensure instruction-following. Design and evaluate prompts using prompt engineering techniques to optimize LLM outputs. Create adversarial prompts to test model failure cases and evaluate factuality of generated content.

Freelance AI Trainer at Appen

Review and rate AI-generated Telugu responses for linguistic accuracy, fluency, and prompt alignment. Evaluate translations based on grammar, completeness, cultural sensitivity, and context.

Associate Analyst – Google Client at Vaco Binary Semantics LLP

Prepare audit reports and quality statistics for human raters on AI evaluation tasks. Automate data handling in Google Sheets using Google Apps Script, improving process efficiency. Contribute to the development of an AI model that analyzes image-topic relevance based on scraped web data. Evaluate AI responses using Reinforcement Learning with Human Feedback (RLHF) under a Human-in-the-Loop (HITL) framework.

Freelance AI Trainer – Coding Expert at Outlier AI

Debug and assess LLM-generated code responses for DSA and full-stack problems (HTML, CSS, JS). Test responses in IDEs like VS Code to identify bugs and ensure instruction-following. Design and evaluate prompts using prompt engineering techniques to optimize LLM outputs. Create adversarial prompts to test model failure cases and evaluate factuality of generated content.

Freelance AI Trainer – Bilingual (Telugu) at Appen

Review and rate AI-generated Telugu responses for linguistic accuracy, fluency, and prompt alignment. Evaluate translations based on grammar, completeness, cultural sensitivity, and context.

Associate Analyst at Vaco Binary Semantics LLP

Prepare audit reports and quality statistics for human raters on AI evaluation tasks. Automate data handling in Google Sheets using Google Apps Script, improving process efficiency. Contribute to the development of an AI model that analyzes image-topic relevance based on scraped web data. Evaluate AI responses using Reinforcement Learning with Human Feedback (RLHF) under a HITL framework.

Freelance AI Trainer at Coding Expert

Debug and assess LLM-generated code responses for DSA and full-stack problems (HTML, CSS, JS). Test responses in IDEs like VS Code to identify bugs and ensure instruction-following. Design and evaluate prompts using prompt engineering techniques to optimize LLM outputs. Create adversarial prompts to test model failure cases and evaluate factuality of generated content.