Mehran Gharooni Khoshkebar

Deployed on-prem LLM services on HPC (SLURM, Singularity, JupyterHub) for secure operations, built session-based access with SQL-backed authentication and admin-level permissions, and established controls (quotas, limits, auditability, usage tracking) to ensure reliability and privacy for 100+ researchers.

Engineered a 24/7 AI customer service bot on Google Cloud Compute Engine; provisioned secure VMs, hardened security, and set up monitoring/alerting for high availability. Created runbooks and operator checklists to support ongoing use, and introduced reliable incident response.

Led the engineering roadmap for an end-to-end AI-first platform (FastAPI backend, PostgreSQL, React/Node.js frontend), shipping production AI automations with on-device setups for Windows/macOS, and delivering integrations via APIs/webhooks (email/calendar, spreadsheets, chat apps). Built reliable real-time and deferred AI workflows with WebSockets and feature flags, introduced robust monitoring, logging, retries, and safe fallbacks, and ensured audit-friendly read-only access to analytics. Led an 8-person cross-functional team, delivering the patent-filed SmartTeach/ATA platform and an internal analytics copilot with automated DOCX reporting. Designed document onboarding flows (PDF/Image ingestion, OCR/multimodal extraction) and QA gates, with clear handover runbooks and training.