VIAM LABS
Independent AI Research Lab

Researching the
next wave of AI

VIAM Labs is an independent AI research lab. We study agentic systems, frontier model behaviour, and AI product design — publishing our findings and building along the way.

viamlabs-research-lab
Running
Orchestrator
Research Agent
Analysis Agent
Synthesis Agent
Output
Experiments Run
1,200+
↑ Active research
Publications
12+
Papers & reports

The lab

Research, built in the open

A small lab working on hard problems at the frontier of AI.

0+
Research Papers
Published papers and technical reports
0
AI Products Built
Research-driven products shipped from the lab
0
Research Areas
Agentic AI, model behaviour, product design, ML
0+
Open Source Tools
Evals, frameworks and libraries released publicly

Research focus

What we study

Our research sits at the intersection of agentic AI, model behaviour, and AI product design — with a focus on problems that matter in production.

Autonomous reasoning & action

Agentic AI

Multi-step reasoning, tool use, and autonomous decision-making in AI systems. We study how agents plan, recover from failure, and coordinate at scale.

  • Multi-agent orchestration
  • Tool-use & function calling
  • Long-horizon task planning
  • Human-in-the-loop systems
Interpretability & alignment

Model behaviour

Understanding what frontier models actually do — interpretability, emergent capabilities, alignment properties, and failure modes under distribution shift.

  • Emergent capabilities
  • Alignment & safety
  • Interpretability methods
  • Evaluation & benchmarking
From research to production

AI product design

How AI-native products are architected, evaluated, and deployed at scale. We study the gap between research results and reliable production systems.

  • AI product architecture
  • Eval & reliability frameworks
  • UX for AI-native products
  • Deployment & observability
Practical methods at the frontier

Applied ML

Bridging research findings and real-world systems — fine-tuning, RAG, retrieval architectures, and inference optimisation for production environments.

  • Fine-tuning & RLHF
  • Retrieval-augmented generation
  • Inference optimisation
  • Data curation & pipelines

Products

What we build

Research-driven products and tools built at the lab — from AI document intelligence to multi-agent infrastructure.

Live
Intelligent document platform

ViamOne

Beyond storage, beyond search. ViamOne's AI understands your contracts, IDs and insurance — extracting insights and turning static files into living intelligence.

Ask AI anythingExpiry & renewal alertsSemantic searchKnowledge graph
Visit viamone.com
Available
Orchestrate AI agents at scale

Agent Runtime

A production-ready runtime for deploying, monitoring, and scaling multi-agent AI systems — built out of our agentic AI research.

Multi-agent orchestrationTool-use frameworkReal-time monitoringAuto-scaling
Learn more
Available
RAG infrastructure at scale

Knowledge Hub

Connect any knowledge source to AI. Secure, scalable RAG infrastructure with semantic chunking and support for 50+ data connectors.

Semantic search50+ connectorsRole-based accessAudit logging
Learn more
Coming Soon
Govern AI access across your org

AI Gateway

Centralised gateway for all AI calls — routing, cost tracking, PII redaction, and compliance guardrails built on Cloudflare's edge network.

Multi-provider routingCost controlsPII redactionCompliance logs
Learn more

Publications

Recent research

Papers, technical reports, and experiments from the lab. All findings published openly.

View all publications

Evaluating tool-calling reliability in agentic pipelines

A systematic study of failure modes when language models invoke external tools across multi-step tasks — with a benchmark suite and mitigation strategies.

AgentsEvalsTool use

Context window utilisation patterns across frontier models

Empirical analysis of how GPT-4o, Claude 3.5, and Gemini 1.5 use their context windows — revealing systematic gaps between stated and effective capacity.

LLMsBenchmarksContext

VIAM-Bench: an open eval suite for multi-agent coordination

An open-source evaluation framework for measuring coordination, task delegation, and error recovery in heterogeneous multi-agent AI systems.

Open sourceAgentsEvals

Join the lab

We're a small team with a long view

We work independently, publish openly, and stay close to the frontier. If you care deeply about hard AI problems, we'd love to hear from you.

Get in touch

Say hello to the
lab

We publish openly and move fast. Whether you want to collaborate on research, ask about our products, or just say hello — drop us a line.

We're open to

Research collaboration
Working with universities, independent researchers, and other labs on shared problems.
Product feedback & early access
Pilots and previews for our products — ViamOne, Agent Runtime, and what's coming next.

By submitting, you agree to our Privacy Policy.