AKTech is a technology startup based in Vietnam, combining research, engineering, and AI to develop solutions for real-world applications. Our work spans software development, intelligent systems, and emerging technologies, with a focus on turning research into reliable, practical products.
Two teams, one loop: research shapes what we build, and products send real data back to research.
Web platforms and AI-powered applications designed for real-world use, combining intelligent technologies, practical software solutions, and user-focused experiences to address diverse needs across industries.
Applied research across machine learning, AI and data-driven technologies, exploring practical solutions to real-world challenges. Every research initiative is an opportunity to drive innovation and create products with real-world impact.
Two flagship projects, both built on our own computer-vision stack. Each began as a research question and is now taking shape with early users.

Studio motion capture needs suits, markers and a dedicated room, so most animators, coaches and clinics never get accurate 3D motion data. Our system tracks people and objects from ordinary multi-camera video and turns their movement into animation-ready skeletons: no suit, no studio. Built for animation and game studios, sports teams and rehabilitation clinics.
Caring for crops row by row is slow, repetitive work, and farms are short of people to do it. Our unmanned ground vehicle drives itself along the rows, uses on-board cameras to tell crops from weeds and obstacles, and handles routine care such as inspection, watering and spraying. Designed for vegetable and fruit farms that want to automate field work without rebuilding their fields.
Our research division works on problems where a better model changes the outcome, and keeps a clear path from paper to product.
Deep learning for medical imaging, starting with chest X-rays: models that give clinicians a faster, more consistent second read, developed with hospital partners and evaluated on real clinical data.
Adapting large vision and language models to Vietnamese data and domain-specific tasks, with a focus on efficient fine-tuning, honest evaluation and safe deployment inside products.
Learning from data that never stops: sensor feeds, video and transaction logs. Our models adapt online and flag concept drift without being retrained from scratch.
Forecasting and decision models for finance and operations: time-series prediction, risk scoring and anomaly detection that turn historical records into forward-looking signals.
Talk directly with our engineers and researchers.
Whether you need a web platform, a computer-vision system or a research partner, tell us about the problem and our team will get back to you.