FlowCore AI (Jan 2026–Present): Manually reviewed, labeled, and categorized 133+ customer support conversations for inte
FlowCore AI (Jan 2026–Present): Manually reviewed, labeled, and categorized 133+ customer support conversations for intent classification, sentiment analysis, escalation triggers, and response quality. Created structured datasets that directly trained the AI to achieve a 70.6% automation rate. Evaluated AI-generated responses for accuracy, tone, safety, and relevance — providing structured feedback for model improvement. Umbraa (2026): Created labeled security datasets for vulnerability classification (CVSS scoring, severity levels, exploitability). Performed AI red teaming — testing autonomous agents for unsafe behaviors, hallucinations, and jailbreak vulnerabilities. Built structured annotation frameworks for security findings. Independent Projects (Oct 2025–Present): Built web scraping agents with automated data cleaning and labeling pipelines. Created labeled datasets for YouTube content categorization, video subtitle timing, and music mood matching.