Introducing 
Express — Go from prompt to pipeline in minutes.
What is Nexla’s Inspiration to Trip multi-agent AI demo at NVIDIA GTC 2026? At NVIDIA GTC 2026, Nexla is demonstrating Inspiration to Trip, a live multi-agent AI travel planner built with NVIDIA and Nebius. The system analyzes a travel video, extracts locations and experiences, enriches them with travel data, and generates a personalized itinerary in seconds. Nexla orchestrates the data pipelines and agent workflows that make the production-scale demo possible.
Creating a controlled AI agent demo is easy; making it reliable, scalable, and functional with real enterprise data is the real challenge. This is the production gap where most AI projects fail, and it’s the gap Nexla is designed to bridge.
At NVIDIA GTC 2026, Nexla is demonstrating exactly that: a live, end-to-end multi-agent pipeline that processes real video input, routes it through governed data products, and delivers a structured output in seconds. Not a prototype. A production-grade system running in front of an audience.
The hardest part of deploying AI agents at enterprise scale is not the model. It is the data underneath it. Enterprise data lives across hundreds of systems, arrives in inconsistent formats, carries governance requirements, and needs to be transformed and routed in real time to be useful to an AI pipeline.
Nexla handles that problem end to end. With 550+ bidirectional connectors, semantic intelligence, support for structured, unstructured data including videos and docs, support for MCP servers, and a conversational data engineering interface in Express.dev, Nexla turns raw enterprise data into governed, agent-ready data products. It also coordinates the multi-agent workflows themselves: routing signals between models, managing context across pipeline stages, and ensuring data quality and lineage at every step.
At NVIDIA GTC, Nexla is also showcasing its implementation of the NVIDIA AI Blueprint for ASR and Video Search and Summarization. Nexla’s pipeline and orchestration product leverages a Ray based framework capable of scaling to millions of files in any major filesystem or object store. The automated data operations accelerates how quickly production AI pipelines can be built and deployed.
“We believe the next frontier of AI is not just generating answers, it is orchestrating intelligent agents that work together to solve real-world problems at real-world scale. This collaboration is a proof point of what becomes possible when best-in-class data integration, foundation models, and purpose-built GPU infrastructure converge.” – Saket Saurabh, Co-founder and CEO, Nexla
To demonstrate this in action, Nexla is running a live multi-agent travel planning workflow at GTC alongside NVIDIA, Nebius, and Tripadvisor to showcase what’s possible for the future of travel planning. The demo, called Inspiration to Trip, compresses the full journey from a social media video to a personalized travel itinerary into seconds.
A user pastes a travel influencer video link into an experimental feature in Tripadvisor’s native AI experience. From there, a five-stage pipeline unfolds:
Every stage of that pipeline depends on Nexla’s ability to connect data sources, enforce quality and governance, and coordinate model interactions reliably. The demo only works because the data layer works.
Watch Demo Video [Here]
Running a multi-agent pipeline live, at demo scale, with real input and real-time output requires more than capable models. It requires infrastructure that does not introduce latency variance, does not degrade under load, and is purpose-built for the workload.
Nexla’s pipelines for this demonstration run on Nebius AI Cloud, which provides dedicated NVIDIA HGX B200 systems with high-performance NVIDIA Quantum InfiniBand networking and a non-virtualized architecture. That matters in practice: shared cloud environments introduce unpredictability that compounds across a multi-stage pipeline. Dedicated infrastructure removes that variable.
The performance difference is measurable. In benchmark testing across video processing workloads at production scale, NVIDIA NIM microservices running on Nebius achieved up to 85% cost efficiency compared to leading cloud-hosted API models, while maintaining comparable output quality.
“Nebius exists to remove the infrastructure barriers between ambitious AI ideas and real production deployments. We are proud to accelerate what Nexla and NVIDIA have built by delivering the specialized infrastructure needed for high-performance, reliable AI systems at scale.” – Laurelle Roseman, VP of Global Partnerships, Nebius
Nexla is at NVIDIA GTC 2026 in San Jose, California, March 16 to 19. The Inspiration to Trip live demo ran at the Nebius booth (#713) on Monday, March 16.
If you missed the live session but would like to see the demo, reach out to info@nexla.com to schedule a personal walkthrough.
Nexla is an enterprise-grade, AI-powered data integration platform that unlocks data from any source into production-ready data products for AI and agents. With support for 500+ connectors and multiple integration styles, including ELT, ETL, streaming, APIs, and agentic RAG, Nexla enables teams to build and manage data flows without writing code. Nexla processes over one trillion records per month for leading organizations across industries and is recognized in over 14 Gartner 2025 Hype Cycles for Cloud and AI. nexla.com.
Nebius, the AI cloud company, is building the full-stack platform for developers and companies to take charge of their AI future, from data and model training to production deployment. Founded on deep in-house technological expertise and operating at scale with a rapidly expanding global footprint, Nebius serves startups and enterprises building AI products, agents, and services worldwide. Nebius is listed on Nasdaq (NASDAQ: NBIS) and headquartered in Amsterdam. nebius.com.
Multi-agent AI is a system where multiple specialized AI models or agents collaborate to complete complex tasks. Each agent performs a specific role, such as video analysis, data enrichment, or reasoning, and passes results to other agents within an orchestrated workflow.
AI agents depend on reliable access to high-quality data. Without governed pipelines, consistent formats, and real-time integration across enterprise systems, AI agents cannot operate reliably in production environments.
Nexla prepares enterprise data for AI agents by connecting sources, transforming formats, enforcing governance, and delivering structured, agent-ready data products. This ensures AI pipelines can access consistent, trusted data at runtime.
At NVIDIA GTC 2026, Nexla is demonstrating Inspiration to Trip, a live multi-agent AI travel planning pipeline. The system analyzes a social media travel video, extracts locations and experiences, enriches them with travel datasets, and generates a personalized itinerary in seconds.
The demo uses a five-stage multi-agent pipeline. A travel video is ingested and analyzed using NVIDIA vision and speech models, Nexla enriches extracted signals with travel data, and NVIDIA reasoning models generate a structured travel itinerary. Nexla orchestrates the data flow and agent interactions across each stage of the pipeline.
Nebius provides the dedicated AI cloud infrastructure powering Nexla’s multi-agent pipeline. Running on Nebius AI Cloud with NVIDIA HGX B200 GPUs and InfiniBand networking, the system delivers the performance and reliability required for real-time AI agent workflows.
Nebius provides purpose-built AI infrastructure designed for large-scale model workloads. Its dedicated architecture reduces latency variability and ensures consistent performance across multi-stage AI pipelines.
NVIDIA provides the foundation models and GPU compute that power video analysis, speech recognition, and reasoning in the pipeline. Nexla orchestrates the data movement and agent interactions that allow those models to work together reliably.
Explore how a multimodal AI pipeline built with NVIDIA models, Nebius infrastructure, and Nexla orchestration converts social media travel videos into structured itineraries.
Agentic RAG systems fail when data is fragmented, stale, or inconsistent. Learn how AI-ready data products with standardized schemas, governance, and retrieval metadata enable reliable, scalable RAG applications.
Raw feeds without context create endless rework. This metadata-first blueprint shows how to turn changing source feeds into governed, reusable data products with automated validation, lineage, and GenAI-ready contracts.
