NEOB Labs · Vienna
NEOB Research & Development
We research, prototype and build the interfaces between humans, businesses and autonomous AI agents - voice, realtime, multimodal and everything that has to run underneath them.
- Research areas
- 8
- Experiments
- 7
- Write-ups
- 23
- In production
- 1
The thesis
One agent. Multiple interfaces. Shared context.
A business should not need four disconnected AI systems for phone, website, chat and messaging. The channel is transport. What decides behaviour - knowledge, memory, business logic, tools, guardrails - belongs in one layer underneath all of them.
One agent definition - instructions, tools, guardrails
The interesting problems are not the channel adapters. They are identity resolution, shared memory under weak evidence, latency inside a synchronous turn, and what happens when a tool call fails halfway. That is what the research areas below are about.
Research
What we are working on
AI Agents
What separates an agent from a chatbot is not the model - it is the loop around it. We work on the control flow, the tool interface and the failure modes that appear once a language model is allowed to act.
02Voice AI
Voice is the hardest interface an agent can have, because the user hears every millisecond of thinking. Our work here is about turn-taking, interruption, and holding conversational state while tools run.
03Realtime AI
Realtime is a systems problem, not a model problem. We study where the milliseconds actually go, which parts of a pipeline can be overlapped, and what latency budget a conversation can survive.
04Multimodal AI
A useful agent rarely gets clean text. It gets a phone call, a photographed invoice, a screenshot and a PDF. We work on how one agent takes in several modalities without collapsing into a router of special cases.
05Computer Use
When there is no API, the user interface is the API. We work on agents that drive a browser or a desktop, and on the containment that has to exist before such an agent touches anything real.
06Agent Infrastructure
Most of the engineering in an agent product is not the agent. It is the runtime, the tool registry, the permission model, the retrieval layer, and the traces you need when something goes wrong at 03:00.
07Conversational AI
Conversation is a transport-independent problem. The same intent arrives by phone, in a web widget or in a message thread - and the differences between those are mostly about time, not about language.
08Omnichannel Agents
A business should not need four disconnected AI systems for phone, website, chat and email. The interesting part is not the channel adapters - it is the shared context and the identity resolution underneath them.
From research to product
Research that has to survive real customers
A lab that only publishes is easy to run. Ours is attached to systems that take live phone calls, which changes which problems count as solved. The route out of this lab runs through prototypes into products.
- 01Research question
- 02Prototype
- 03Experiment write-up
- 04Engineering
- 05Production system
Production system
bitpull.ai - AI agents for customer communication
A production SaaS platform that lets businesses deploy AI agents for customer communication: an agent answers the phone, handles the website widget by voice or text, books appointments against a real calendar and hands over to a person when it should. It is the system most of the voice, realtime and tool-use research here feeds into - and the reason our latency and reliability constraints are not hypothetical.
- Live Phone (SIP, own numbers)
- Live Web widget - voice and text
- Live Messaging integrations
- Research Email as an agent channel
Latest
From the lab
Reading what the customer actually sent
Customers do not send clean text. They send a scanned price list, a photographed sign and a PDF exported from a word processor in 2011.
ExploratoryMCP as the tool boundary for a multi-tenant agent platform
MCP standardises how tools are described and transported. It does not standardise who is allowed to call them - and in a multi-tenant product that is most of the work.
ExploratoryA browser agent for portals that have no API
Reading from portals was tractable. Writing to them was not - and the honest conclusion was to stop trying to make it autonomous.
Evaluating voice agent quality
Word error rate tells you about the transcriber. It tells you almost nothing about whether the caller got what they rang for.
Lab noteDesigning human-in-the-loop AI systems
Adding a human is easy. Adding one whose attention is actually worth something is the design problem.
The NEOB domains
Three domains, three jobs
This lab
Research, experiments, engineering write-ups and prototypes. Technical, English, and happy to say when something did not work.
The company
NEOB Technology GmbH - AI solutions, enterprise AI, consulting and delivery. Who we are and what we do for clients.
bitpull.aiThe product
The commercial platform: AI voice and chat agents for customer communication, with pricing, industry pages and a live demo.