TRAVIS: NYC DOE Conversational AI Assistant
NDA2022 to present · NYC Department of Education, Office of Pupil Transportation & District Programs
Production AI assistant that turns plain English staff questions into live, permission scoped data queries and workflows.
Architecture and outcomes are shareable at this level of detail; internal prompts, code, and proprietary implementation details are restricted under NDA.
TRAVIS is a production conversational AI assistant for NYC DOE's Office of Pupil Transportation. It turns plain English staff questions into live, permission scoped data queries, reports, and multi step workflows, cutting certification processing time by over 20%.
The problem
Before TRAVIS, staff had to navigate multiple systems and manual reports to answer routine operational questions, slowing down certification and day to day decisions.
System design
TRAVIS request flow
Process & prototyping
- Two stage LLM agent: intent and domain classification, then schema validated query generation with Pydantic.
- Async Python 3.11 and FastAPI backend with 15 endpoints, migrated from an earlier Node.js implementation.
- Retrieval augmented query planning: few shot retrieval plus a semantic query plan cache over Qdrant vector embeddings, cutting redundant LLM calls.
- Layered production guardrails: NeMo Guardrails, PII masking, and prompt injection defense.
- JWT authentication, role based access control, and a Redis backed permission cache.
- Containerized with multi stage Docker builds.
Outcome & impact
TRAVIS is live in production and cut certification processing time by over 20% for NYC DOE staff.
Role & collaboration
Team
Built as part of Kshitij's ongoing NYC DOE engagement via Athreya Inc.