Pioneering intelligent solutions for a smarter world
Discover Our Visionstm.ai is a leading ai consultancy, research, and development business. We specialize in cutting-edge technologies and solutions across various industries.
Our expertise spans multiple domains, including:
With a team of experienced researchers and developers, we're committed to pushing the boundaries of AI and ML to solve complex problems and drive innovation.
Expert guidance on implementing AI solutions for your business, tailored to your specific needs and industry challenges.
Custom machine learning models and algorithms designed to solve your unique problems and optimize your processes.
Cutting-edge research in AI and ML technologies, pushing the boundaries of what's possible in various domains.
An open-source data science Python toolkit for Parkinson's Disease, supported by the Michael J. Fox Foundation. Our team was involved in designing, developing, and validating a comprehensive PD data science toolkit.
The project aims to develop robust clinical outcome measures for therapeutic trials and support longitudinal investigations of disease mechanisms through the analysis of data collected primarily from wearables and smartphones.
Innovative IT system for the development of commerce, integrated with intelligent assistance. Our team led the development of advanced AI features for this innovative commerce platform.
We implemented a voice interaction interface using Mozilla's open-source algorithms for real-time speech recognition in Romanian, developed natural language processing modules for intelligent product associations, and created image and facial recognition systems for enhanced user experiences. We also implemented a machine learning-based pathfinding algorithm using Reinforcement Learning to guide users to desired products efficiently.
At GoodMine, we developed cutting-edge algorithms using open-source large language models. Our work focused on identifying product categories and states from descriptions.
We created an LLM-controlled web crawler for populating vector databases and refined data collection processes using techniques like DPO (Direct Preference Optimization) and RLHF (Reinforcement Learning from Human Feedback).
As researchers for the Science Museum Group, we contributed to "Me, Human," a live scientific experiment investigating how traits from our 500-million-year-old vertebrate brain underpin uniquely human behaviors.
This project, part of the Live Science initiative at the Science Museum, involved carrying out research using museum visitors as volunteers, bridging the gap between scientific research and public engagement. We focused on behaviors like face recognition and speech generation.
At Entrepreneur First, we worked on developing a company that leveraged state-of-the-art deep learning algorithms in automated speech recognition (ASR), natural language processing (NLP), and natural language understanding (NLU). The project aimed to reduce the need for extensive meetings in the workplace by automating certain communication processes, utilizing AWS and Google Cloud infrastructures for training and deployment.
At myrecovery, we developed an in-house prototype desktop application using PyQt. This innovative app extracted various 2D gait features from smartphone videos using Convolutional Neural Networks (CNNs), segmentation, and pose estimation techniques. The project involved setting up an experimental data collection protocol in a gait detection lab, which was crucial for the pose estimation training pipeline.
For Atlantis Capital Management, we developed a bespoke Python/SQLAlchemy market intelligence tool designed to capture, store, and extract knowledge from market trading activity. This project also involved implementing financial forecasting algorithms and state-of-the-art machine learning techniques, which were back-tested on market data to ensure reliability and accuracy.