A London practice working at the join between large-scale product engineering and applied AI. Small, senior, hands on the keyboard.
Multi-agent and single-agent systems with explicit stopping conditions, escalation thresholds and human handoff. Designed so the failure mode is known before it ships.
Model Context Protocol servers that expose existing systems as tools a model can actually use, with tool descriptions and error responses written for self-correction.
Eval suites, LLM-as-judge methodology and quality thresholds, so an AI feature can be held to a standard rather than a demo.
Automation across the tools a business already runs, built to be maintained by the team that inherits it.
NOION is led by Jagadesh Thondapu (Jag), a founding team member at Reflexis Systems, where he was Head of Technology through the company's growth and its acquisition by Zebra Technologies in a transaction valued above $500M. He went on to lead the Workforce Optimisation portfolio at Zebra as Director of Product.
That history means the work here starts from how enterprise software is really built, operated and bought, not from a model demo. The domain we know best is frontline and deskless workforce software: retail, hospitality, field operations, and the systems that schedule and direct people who do not sit at desks.
Engagements start with a conversation about the problem, not the technology.