Research for a more human-centered world

Agentic Systems
for a More Human Future

We advance fundamental research and build real-world systems where artificial intelligence, intelligent systems, and human-centered applications meet — with a focus on healthcare, multi-agent collaboration, research infrastructure, and knowledge systems.

Lab facts

Research areas

From fundamental research to real‑world impact.

Artificial
Intelligence

Foundation models, reasoning, and human-centered AI.

Intelligent
Systems

Multi-agent systems, human–AI collaboration, and complex systems.

Multi-agent
Systems

Runtime safety, coordination, and verification for systems with multiple agents.

Healthcare AI

AI for health, biomedicine, and clinical decision support.

Knowledge
Systems

Knowledge representation, scientific AI, and information systems.

Secure AI
Bots

Employee-facing AI bots, approved runbooks, separated consents, and action receipts.

Featured research programs

Long-term programs building toward AI that is more capable — and more beneficial.

Microscopic view of botanical leaf cells from medicinal plant research

AIMed

AI for healthcare and Vietnamese traditional medicine.

Applied healthcare AI that puts medical knowledge and intelligent assistance to safe, real-world use.

5,190+ medicinal herbsInteraction checkingAI plant recognition

Public platform — live

Explore the AIMed program
A research visualization of software agents pausing at a human approval checkpoint

ActantOS

Runtime safety and verification for multi-agent systems.

A self-hosted enforcement kernel that intercepts tool calls between agents and the systems they act on — evaluating each call against policy before allowing, pausing for approval, or blocking.

Fail-closed enforcementCedar policy engineOne-use human approvals

v1.2.0 “Quiet Open-Core” — shipped, self-hosted

Explore the ActantOS program
Editorial illustration of a service-desk bot answering from an approved runbook while every action is filed into an audit ledger

ActantBot

Security and runtime-safety agents on the ActantOS kernel.

The WorkDesk bot family for employee IT support: guided troubleshooting from approved runbooks, requests filed through three separated consents, and security reports taken in a private channel — every write decided by the ActantOS kernel.

WorkDesk bot familyApproved runbooksThree separated consents

WorkDesk — shipped with ActantOS

Explore the ActantBot program
A university research commons with scholarly books and shared data resources

ScienzaOS

Research infrastructure for open science and collaboration.

The platform for academic life: Research Agent, Teaching Studio, and Business Collaboration on one shared foundation of sources, review, rights, and provenance — agents prepare, people retain authority.

Research AgentTeaching StudioBusiness Collaboration

Pre-launch — access by request

Explore the ScienzaOS program
Illustration of a request moving through a policy gate and human approval checkpoint to an audit ledger

Agent Platform

Governed AI agents with clear permissions and an auditable record of every action.

An operating layer that connects channels, agents, people, and systems — enforcing rules and scoped permissions, pausing for human approval when needed, and recording every action and outcome.

Scoped permissionsHuman approvalsAction records

In development

Explore the Agent Platform program
Editorial illustration of a scanned document being transformed into structured data while a researcher reviews the source

AgenticOCR

AI-assisted document-to-data workflows with human review.

Turns scanned documents into structured data, keeps source pages beside extracted results, and supports review, editing, and export in one workflow.

Document-to-data workflowPage-level source reviewJSON, Excel, HTML export

Staging showcase — live

Explore the AgenticOCR program

From research to reality

From research question to working system

Research at ERA Lab does not stop at a paper or a prototype. Each program moves through the same 5 stages, and every stage is published where it can be checked.

  1. Research question

    What should an autonomous system be allowed to do, and who decides?

    ActantOS
  2. Methods

    Runtime enforcement, policy evaluation, evidence lineage, and measurement — the engineering that makes an answer testable.

    ActantOS v1.2.0
  3. Working systems

    Five long-term programs turn those methods into software that runs: platforms for healthcare AI, agent safety, open science, governed agents, and document understanding.

    Projects
  4. Verification before assertion

    A public platform, a shipped release, a published record, or a documented source — every claim on this site resolves to one of them, and in-development work is labelled as such.

    Research records
  5. Intended impact

    Agentic systems that hospitals, universities, and research teams can adopt without giving up control or accountability.

    Collaborate

Selected publications

Recent research from the lab.

View all research
  1. 2026 · Journal article

    Data driven adaptive SVC control for 110 kV substations using machine learning and edge deployment validation

    Nguyen Thanh Quang · Ngo Trung Kien

    DOI · Discover Electronics
  2. 2026 · Conference paper

    Multi-Agent Systems for Nontraditional Financial Security: Architecture, Mechanisms, and Policy Implications

    Trung-Kien Ngo · Thu-Huyen Hoang · Thanh-Tung Trinh

    HSB faculty publication record
  3. 2026 · Conference paper

    Smart and Sustainable Agriculture: Offline Reinforcement Learning for Intelligent Handover in Hybrid VLC/RF Agricultural IoT Networks

    Ngo Trung Kien · Vuong Thi Nhung · Le Anh Ngoc

    Springer · Proceedings of CSA 2025
  4. 2026 · Journal article

    Circularity and Sustainability of Rice Straw Biomass Supply Chain Assessment for Bioenergy Production: Case Study in Viet Nam

    Nguyen Thanh Quang · Nguyen Ngoc Bach

    DOI · BioEnergy Research

Patents

Recent patent records from ERA Lab members.

A selection of recent patent entries in LiFi, smart lighting, positioning, and IoT.

View patent records
  1. CV 1.1 2025 Vietnam

    Method for measuring the distance between two lamps with a camera-equipped smartphone and setting lamp coordinates in a lighting system to provide LiFi positioning services

    Vietnam Intellectual Property Office

    Patent / application number: VN1-0053734-000

    Inventors / authors: Nguyễn Văn Tràng

    Search the official Vietnam patent database
  2. CV 1.2 2025 Vietnam

    Light-signal modulation method for virtual lamp models and virtual smart-lighting systems in 3D simulation, with lamp selection and control through augmented reality

    Vietnam Intellectual Property Office

    Patent / application number: VN1-0053706-000

    Inventors / authors: Nguyễn Văn Tràng

    Search the official Vietnam patent database
  3. CV 1.3 2025 Vietnam

    Dual-mode WiFi-LiFi smart lamp, data-routing method, and antenna-switching process

    Vietnam Intellectual Property Office

    Patent / application number: VN2-0004110-000

    Inventors / authors: Nguyễn Văn Tràng

    Search the official Vietnam patent database
  4. CV 1.4 2025 Vietnam

    System and method for controlling lighting devices based on the Internet of Things

    Vietnam Intellectual Property Office

    Patent / application number: VN1-0053216-000

    Inventors / authors: Nguyễn Văn Tràng

    Search the official Vietnam patent database

Our people

An interdisciplinary team of researchers, builders, and partners.

General contactcontact@eralab.sihello@eralab.si

Meet the team

Selected member work

Teaching, research, and applied work.

An open research report with data visualizations and microscopy figures
“AI is a general-purpose technology. Our goal is to make it more useful, more reliable, and more human-centered in the worlds that matter most.”
—  ERA Lab

News

News and notes from the lab.

All news

2026 · 10

ActantBot WorkDesk documents governed employee IT support

WorkDesk documents employee IT support bots that answer from approved runbooks, route requests through separated consents, and return receipts from the ActantOS kernel.

Researchers sharing notes around a table

Collaborate

Research is a collective effort.

We work with researchers, clinicians, industry partners, and public institutions to build AI that is trustworthy and beneficial.

Discuss a collaboration

Who we work with

Open to collaboration across disciplines.

  • Universities
  • Research institutes
  • Hospitals and healthcare organizations
  • Industry partners
  • Open science community