Mohsin Hayat

Built production conversational AI experiences for airline clients including World2Fly and LASER Airlines across Python services, LLM agents, and airline booking workflows. Integrated conversational agents with airline systems and KIUPSS to support booking information, payments, check-in, flight changes, ancillary services, and passenger support.

Built and shipped a production conversational NL-to-SQL agent used by leadership, managers, and marketing teams to query campaign data across 30+ tables in plain English, reducing dependency on manual SQL and cross-schema discovery. Engineered a safety layer using schema-aware retrieval, table allowlists, date-bounded queries, structured outputs, and validation checks to reduce unsafe/invalid/expensive query generation. Architected a budget allocation copilot using a 3-level hierarchical multi-agent system with Graph reasoning over ~$5M/month in ad spend, with business-friendly explanations for non-technical stakeholders. Built an AI ingestion pipeline (AWS SNS→SQS→ClickHouse) and implemented long-running (~1 hour+) Graph pipelines for recommendations.

Owned architecture and development of core orchestration/back-end APIs (orchestration, task, design). Scaled systems to 100K+ monthly workflows, contributing to growth to $1M+ ARR. Improved CI/CD pipelines using Docker/Kubernetes and embedded TDD with 100+ e2e tests (Cypress), reducing deployment time by ~50% and increasing release velocity. Designed and implemented event-driven architecture (CQRS) that streamlined AI integration with RAG.