Architected a secure natural language interface for government datasets using Gemini-3.5-Flash and DuckDB, implementing a Pydantic-validated SQL generation pipeline that eliminated manual query writing and ensured 100% adherence to read-only data constraints via automated keyword blocklisting.

• Optimized query performance with an embedded DuckDB engine over CSV backends, reducing data retrieval latency by 3X.
• Enabled real-time visualization of multi-million record datasets through a custom FastAPI and Plotly integration.
• Developed an Audit-First architecture by integrating a persistent SQLite-based logging system.

Er1s.AI

Architected an enterprise-grade AI Visibility & Citation Tracking platform using Node.js, Express, and PostgreSQL, leveraging Google Gemini to benchmark brand presence, citation frequency, and competitor share-of-voice across dynamic LLM search contexts.

• Engineered a high-performance, asynchronous scanning pipeline using Redis and Bull MQ to handle concurrent LLM requests, reducing processing latency by 60% via parallel batching.
• Designed a robust mention-detection and scoring engine utilizing Prisma ORM to map, rank, and store brand sentiment metrics from thousands of unstructured AI responses.
• Built a premium Glassmorphism analytical dashboard featuring an interactive floating console for query-level win/loss drilldowns and real-time visualization..