Mahesh Kale

I've built AI applications including a groundwater conversational chatbot, an automated data science platform, and a RAG-based video assistant. My work combines LLM prompt engineering, SQL, vector search, and Python to turn complex data into useful answers.

For the groundwater assistant, I translated water-level data into plain-language responses for 50+ non-technical users, reduced average query response time by 80%, and tested 100 edge-case scenarios to improve answer reliability.

On an AI-powered data science platform, I automated data cleaning, visualization, and business-strategy generation for datasets of 10,000+ rows and columns. I reduced manual data-preparation effort by approximately 90% through self-healing pipelines and automated data operations.

During my Artificial Intelligence internship at XTRAGRADE, I created AI architecture proposals and built Python automation that cut completion time from 30 minutes to 5 minutes. I also used structured prompting with GPT-4, GPT-3.5, and Claude to reduce off-target outputs by 60% across 90 test queries.