Murali Manohar

I'm building production AI solutions at Meril Life Sciences for patent intelligence, including end-to-end NLP pipelines that process unstructured patent data through KTmine APIs, semantic chunking, metadata mapping, Streamlit, and MinIO.

I developed computer vision workflows for automated diagram label extraction using Florence-2 VLM and EasyOCR, while evaluating Qwen2.5-VL and Llama3.2-Vision alternatives. I also built patent analytics and topic-discovery modules using K-Means, HDBScan, TF-IDF, BERT embeddings, Plotly, NetworkX, and Power BI.

My work includes a scalable evaluation framework for patent-focused GenAI agents, with synthetic test cases, LLM-as-a-Judge evaluation, evidence-grounded assessment, FastAPI, and PostgreSQL.

Previously, as a Machine Learning Engineer Intern at Mergen IT, I developed classification and regression models and built reliable ML pipelines through EDA, data cleaning, feature engineering, hyperparameter tuning, and SQL. I also created an agentic misinformation detection system and a railway safety monitoring system that achieved over 90% F1-score.