Fyul
AI-powered quality control monitors the print manufacturing pipeline, detecting print defects with 95–99.9% accuracy.
Clients
UPB
A construction-site monitoring project using autonomous quadruped robots and high-resolution LiDAR to create BIM-aligned 3D scans, track progress, and support 24/7 security.
Twino
Optimized risk-analysis processes and developed new models to support more effective risk assessment.
Alarm.com
Research and development for drone-control systems for autonomous navigation and security applications.
Parkexpert
AI-powered number plate recognition system that works with existing cameras, tracks vehicles across frames, and reconciles sightings across cameras to create accurate parking sessions.
Daugauvipls slimnīca & Tet
Development of an AI assistant for the medical field, powered by large language models and a RAG system and tailored to the Latvian language, to improve the quality of healthcare services.
Aspired Latvia
Tailored AI training that equips teams with practical, business-ready skills through hands-on learning designed around their needs and workflows.
Centrālas statistikas pārvalde
Implementation of a pilot project for data classification and automation.
LIAA
An AI solution that automates email drafting, supports employee learning through a training agent, and equips staff with practical skills to use AI effectively and responsibly.
Latvija.lv
A unified, accessibility-first design system for a national public services portal, aligned with the European Accessibility Act.
Weedbot
Precision laser-weeding system that identifies weeds with up to 95% accuracy for efficient, sustainable crop management.
Watersons
Novel model to predict biological contamination risks in drinking water using standard sensor data.
Catchbox
Partnered with a leading regional microphone manufacturer to develop deep-learning models that remove feedback, typing, traffic, background voices, and mechanical noise from audio in real time.
Sportacentrs
An AI vision model that detects and rectifies partially obstructed tabloids into orthogonal projections, achieving over 95% accuracy with fewer than 500 training samples.
Zippyvision
A high-precision AI inspection system for bandsaw operations, detecting defects in wooden planks with 70–99% precision to support automated cutting and manufacturing planning.
Chromesword
Novel model solvent-gradient optimization for analytical chromatography, automating experiments to separate unknown compounds and reducing development time from two weeks to two hours.
Scope technologies
AI-powered damage detection for rental-car inspections, segmenting scratches, dents, chips, dirt, and rust from smartphone photos—with 0.88 IoU despite challenges such as distinguishing dents from reflections.
Rietumu Banka
A cross-department innovation workshop focused on identifying practical opportunities to adopt AI, improve workflows, and address shared business challenges.
Saeima
Bipartisan training sessions helping members of parliament use LLMs for legislative analysis, critically evaluate AI-generated answers, and understand industry risks.
MILA
Branding development for Latvian association of artifical intelligence.