Key Developments:
DoD Solution and Axelera AI develop onboard AI computing for high-speed interceptors capable of autonomous targeting and guidance
General Chereshnya and Alta Ares plan to integrate AI-powered guidance and computer vision into the Bullet anti-Shahed interceptor
MacPaw partners with Liquid AI to build an on-device AI stack for macOS, combining local foundation models with cloud access
!FEST spins out its internal data team into gloozd, a standalone Data & AI agency targeting $1 million in profit as its first financial milestone
An Odesa-based developer runs a 28.9 million-parameter language model entirely on a sub-$10 ESP32-S3 microcontroller
DEFENSE AI & AUTONOMOUS SYSTEMS
DoD Solution Partners with Axelera AI on Edge AI for High-Speed Interceptors
Estonian-Ukrainian defense technology company DoD Solution has partnered with Netherlands-based Axelera AI to develop onboard AI computing systems for high-speed autonomous interceptors. DoD Solution will embed Axelera AI’s Metis M.2 module in its AURA AI autonomy platform to provide sensor processing, target recognition and tracking, navigation, and guidance without relying on external communications or computing infrastructure. The companies plan to test and validate the technology before moving toward commercial deployment, with potential applications across other autonomous defense platforms.
Read more
General Chereshnya and Alta Ares to Integrate AI Guidance into Bullet Interceptor
Ukrainian defense manufacturer General Chereshnya signed a memorandum with French defense technology company Alta Ares to integrate AI-based guidance into its Bullet anti-Shahed interceptor. The system will combine Alta Ares’ Pixel Lock terminal guidance and Cyclop C2 technologies, using computer vision and multi-sensor data fusion to track targets in GPS/GNSS-denied or electronically contested environments. Alta Ares recently closed a €50 million Series A round, while General Chereshnya says it supplies more than 200 Ukrainian military units and has a production capacity of up to 100,000 drones per month.
Read more
AI PRODUCTS & PLATFORMS
MacPaw Partners with Liquid AI to Bring Local AI Models to macOS
Ukrainian technology company MacPaw entered a long-term strategic partnership with U.S.-based Liquid AI to develop an AI stack capable of running foundation models directly on Apple computers without a constant internet connection. The architecture will combine Liquid Foundation Models with MacPaw’s Elix inference engine and Mnemos memory infrastructure, with its Eney assistant becoming the first product to adopt the technology. Most processing will run locally on Apple Silicon to improve response times and data privacy, while cloud models will remain available when needed. MacPaw also plans to make the infrastructure available to third-party developers through Setapp after initial deployment in Eney.
Read more
BUSINESS & ENTERPRISE AI
Ukrainian Hospitality Group !FEST Launches Data & AI Venture gloozd
Ukrainian hospitality group !FEST spun out its internal data team into gloozd, a standalone Data & AI agency focused on data infrastructure, custom AI development, and customer analytics. The company has more than 15 employees and has completed over 20 projects, including an AI assistant for veterans; the amount invested in the spinout was not disclosed, although management said initial launch costs had already been recouped through proof-of-concept work. gloozd is targeting long-term partnerships rather than one-off projects and has set $1 million in profit as its first financial milestone.
Read more
EDGE AI & DEVELOPER ECOSYSTEM
Ukrainian Developer Runs 28.9M-Parameter AI Model on Sub-$10 Microcontroller
An Odesa-based Ukrainian developer known as slvDev ran a 28.9 million-parameter language model on an ESP32-S3 microcontroller costing under $10, using 4-bit quantization to reduce the model to 14.9 MB. The model operates fully on-device without cloud connectivity and generated text at about 9.88 tokens per second in testing, demonstrating the potential for low-cost edge AI on highly constrained hardware. The open-source ESP32-AI project has attracted more than 3,500 GitHub stars and over 440 forks.
Read more
