A prototype is not a system.
Production work has to account for real inputs, edge cases, access, failure, and the people responsible for the outcome.
Westland Labs works in that distance: where architecture meets real data, edge cases, cost, ownership, and the people accountable for the result.
PhD, Carnegie Mellon University. Published researcher in AI and natural-language processing. Manuel's work brings research discipline to practical questions about how an AI system should act, how teams should evaluate it, and where human judgment must remain.
When you work with Westland Labs, you work directly with the person who designs the system. No account managers. No hand-offs.
Production work has to account for real inputs, edge cases, access, failure, and the people responsible for the outcome.
The team agrees how to evaluate quality, reliability, and cost before treating a promising result as an operating capability.
We don't build black boxes. Every system includes documentation, handover sessions, and architecture decisions you can understand, own, and extend. Your team gains the capability, not just the tool.
Deep specialization in multi-agent architectures: agent coordination, adaptive reasoning, tool orchestration, MCP and A2A protocol integration, persistent memory systems. Production experience with LangGraph, LangChain, LlamaIndex, and Claude Agent SDK across real deployments.
Alongside agents: NLP pipelines, recommendation systems, and ML platform engineering. Evaluation frameworks, model governance, inference cost management, observability. Built and shipped in Python across AWS and GCP.
Based in Dublin, Ireland. Working with clients worldwide.