TanviDubey

Deliver production AI platforms end to end: architecture, data modeling, frontend, backend, and deployment across marketing ops, agent orchestration, AI security, and healthcare.

• Designed multi-tenant SaaS architecture with org-scoped data isolation, RBAC, invite-based onboarding, and usage-metered Stripe billing, serving admin and tenant portals from one codebase.
• Engineered RAG pipelines over enterprise and clinical knowledge bases with ChromaDB/FAISS vector search and retrieval tuning, raising answer accuracy 80% over a keyword baseline.
• Built multi-agent orchestration with tool-calling into Slack, MS Teams, Jira, and HubSpot, splitting a Node/Express control plane from a Python FastAPI runtime executing agent steps.
• Optimized LLM serving for 50% faster inference and trained a YOLOv8 detector to 95% accuracy, cutting latency and compute spend.
• Containerized services with Docker and shipped to AWS (ECS, ECR, S3) and Vercel behind Nginx, with JWT auth, Fernet encryption, Socket.IO real-time updates, and Vitest/Playwright coverage.

Applied data science and AI to community product workflows, from forecasting engagement to automating editorial tagging.

• Built predictive models on community engagement data reaching 85% forecasting accuracy, backed by EDA and feature engineering.
• Automated content tagging with a Perplexity AI classifier returning structured JSON, removing manual editorial triage.