At IDeaS Revenue Solutions, I added LLM-written commentary to a Python pipeline that compares forecasts with projections and renders reports in three formats. I also created a Streamlit analytics app that turns Excel exports into rate trends and pricing-decision checks.
I built Proofline to detect numerical hallucinations in LLM-written reports. Its evaluation used 32k labeled claims across three domains, reaching 98.9–100% recall on unseen phrasing.
I co-built a real-time AI voice sales agent that qualifies leads by phone. Model benchmarks, cache-friendly prompts, and streamed playback helped bring median response latency to about 2.9 seconds.
Earlier at IDeaS Revenue Solutions, I migrated an enterprise revenue-management dashboard to Angular and Spring Boot, and streamlined customer-care email tracking with Power Automate. At IFM Engineering Pvt. Ltd., I identified and mitigated XSS, SQL injection, and misconfiguration risks in internal security audits.
