Core contributor to ALOHA (adaptive deep learning on heterogeneous architectures), TeamPlay (energy-aware multi-core platforms), and AIDA (AI data analysis for space).
SYSTHMATA YPOLOGISTIKIS ORASHS IRIDA LABS AE
Greek SME specializing in computer vision and deep learning optimized for embedded, energy-efficient hardware across manufacturing, space, and IoT.
Their core work
IRIDA Labs is a Greek SME specializing in computer vision and deep learning, building intelligent visual recognition systems that run on embedded and heterogeneous hardware platforms. Their work spans industrial manufacturing (additive/subtractive machine vision), space data analysis, and IoT edge computing. They develop AI inference solutions optimized for energy-efficient, real-time deployment — bridging the gap between powerful neural network models and resource-constrained devices.
What they specialise in
Participated in both BOREALIS and Symbionica, which developed reconfigurable machines for additive and subtractive manufacturing of advanced components.
Contributed to AIDA, applying artificial intelligence techniques to space data analysis — a natural extension of their visual recognition expertise.
ALOHA and TeamPlay both address deploying computation on constrained, distributed devices — IoT endpoints and multi-core embedded systems.
Early participation in GHOST explored Galileo satellite navigation as infrastructure for smart city services.
How they've shifted over time
IRIDA Labs started in 2015 with hardware-oriented projects — smart city navigation (GHOST) and machine vision for advanced manufacturing (BOREALIS, Symbionica). From 2018 onward, their focus shifted decisively toward deep learning, convolutional neural networks, and AI deployment on heterogeneous computing platforms (ALOHA, TeamPlay, AIDA). This trajectory shows a company that moved from applied vision systems in industrial settings to becoming a specialist in efficient AI inference — making neural networks run fast on low-power hardware.
IRIDA Labs is converging on AI model optimization for constrained devices — expect them to pursue projects in edge AI, TinyML, and embedded inference across multiple application domains.
How they like to work
IRIDA Labs operates exclusively as a project participant, never coordinating — typical for a focused technology SME that contributes specialized components rather than managing large consortia. With 57 unique partners across 14 countries in just 6 projects, they work in medium-to-large consortia and have built a broad European network. This wide partner base suggests they are adaptable and easy to integrate into new teams, though they haven't established deep recurring partnerships.
IRIDA Labs has collaborated with 57 distinct partners across 14 countries, a remarkably wide network for an SME with 6 projects. Their reach is solidly pan-European, with no visible concentration in a single region beyond their home base in Greece.
What sets them apart
IRIDA Labs sits at a specific intersection that few SMEs occupy: they combine deep computer vision expertise with hands-on experience deploying AI models on energy-constrained, heterogeneous hardware. Their project history shows they can apply this skill across very different domains — from factory floor inspection to satellite imagery. For consortium builders, they offer a rare package: a small, agile company that understands both the AI algorithm side and the embedded systems reality of making it work on actual devices.
Highlights from their portfolio
- ALOHATheir largest-funded project (EUR 357,500), directly aligned with their core mission of running deep learning models efficiently on heterogeneous architectures — the clearest expression of what IRIDA Labs does.
- TeamPlayTheir single highest EC contribution (EUR 430,625), focused on time, energy, and security analysis for multi-core platforms — showing their value in performance-critical embedded computing.
- AIDADemonstrates cross-sector versatility by applying their AI and computer vision capabilities to space data analysis, opening a new application domain.