For founders pursuing commercial opportunities predicated on the application of machine learning and artificial intelligence.
Adaptive Pulse is a SaaS company that analyzes a company’s unstructured data, such as their communications with customers, to predict churn, expansion and referral signals, conversions, and more. These predictions help organizations prioritize retention efforts and increase revenue.
AI Health Highway is developing AiSteth, a smart stethoscope to screen, detect, and predict heart and lung disorders using AI and ML. The device enables primary-care clinics and other healthcare staff to monitor heart/lung patterns, and detect diseases before they strike.
Areto Labs is developing automated content moderation and delivery software to help companies and organizations protect their public figures and employees online while increasing brand loyalty and engagement. The company uses natural language programming (NLP) for sentiment analysis and microaggressions detection, to moderate online abuse, and deliver culture campaigns through automated messaging.
Arkangel AI has developed Hypocrates AutoML, a no-code platform that allows healthcare professionals to train and deploy disease detection models in minutes. The venture is creating white label models for pharmaceutical companies, hospitals, and insurance companies across the Americas and Europe.
Haloo is a self-serve online trademark search and registration platform for small business and enterprise clients that integrates sophisticated legal thinking with AI. They offer instant professional trademark searches, fail-safe trademark applications and automated brand enforcement. This makes the complex process of trademarking faster, more accurate and more affordable.
Lux Aerobot is a space robotics company specialized in the design, manufacturing, and operation of high-altitude platforms (HAPs) for Earth observation. Leveraging its own unique data set, Lux is building tools to support decision-making ranging from wildfire and coastline monitoring to national security.
Mely.ai is developing a cloud-based, SaaS, AI engine to help supply chain companies and logistic industries extract data automatically from documents. The company uses computer vision and natural language processing to achieve 90 percent time saving and 80 percent labour-cost reduction on back-offices. The product also provides intelligent process automation and business intelligence based on the extracted data.
Memorable is a software that predicts the memorability and saliency of images and videos, as well as elements within them. This technology developed at MIT increases ROI for marketing and design professionals, especially in the awareness space, where predictive solutions are very scarce.
Sesh provides AI-powered coaching to help parents build healthy relationships, manage behavioural issues, and make better decisions. Using multi-modal analysis, including video, audio, and text, Sesh offers a personalized parenting app in addition to interactive courses to support parents. Parents of young children looking to help prevent tantrums are the initial target audience for the app.
Simmunome builds AI-driven computational disease models for use in clinical R&D. The venture focuses on accurately replicating biological processes to simulate clinical trial outcomes without the need for a client to step into a lab. The venture has two validated models: for Alzheimer’s disease and melanoma. Simmunome’s target markets span big pharma, small/mid-size biotech, and academia.
Visual Behavior is developing an artificial visual cortex software emulating the human brain to provide robots with a high-level understanding of their surroundings. The venture allows robot manufacturers to target new markets that are presently inaccessible: autonomous robots and drones, complex action cobots (industry), indoor human assistance, and advanced human-robot interactions.
Biotech Square has developed a web-based platform that allows drug development teams in the pharmaceutical industry to query large regulatory databases for context-specific regulatory precedent. Leveraging natural language processing (NLP) technology, the platform provides users with entire document sections to answer specific questions on regulatory precedent (why a drug was approved, what concerns the FDA had with similar products) in a few clicks rather than searching manually through documents. Accessing more and higher quality precedent records allows drug development teams to improve their regulatory strategies and regulatory approval submission package.
Shakudo has developed an MLOps platform called Hyperplane. Hyperplane connects to existing MLOps tools and is an environment that enables data scientists to independently code, run experiments, and automate deployment of their models without requiring downstream engineering and DevOps support. This increases the speed to deployment, and reduces the need to collaborate with large and resource-intensive engineering and DevOps teams in order to get models into production.
TrojAI software simulates real-world adversarial attacks (natural and malicious attacks) on AI models to monitor model or dataset security during training. As a result, TrojAI provides insights into the weaknesses of the AI model and dataset, enabling users to know precisely howt to alter their model or dataset to achieve better performing and more secure AI models. This enables users to identify points of vulnerability in their model and data set before deployment and thus reduces the time needed to retrain the model for failures that would occur post-deployment.
Waverly has developed a content recommendation engine that gives individual users the ability to edit their own recommendation algorithm using written paragraphs instead of code. This allows users to have transparency and control over what content is recommended to them. The first iteration of the product is a mobile app that allows users to develop their own content recommendation algorithm and connect with other users to share and follow each other’s content recommendation algorithms as well.
Advai tests the robustness of AI models using adversarial AI that manipulates inputs to AI systems in a way that is invisible to humans, but causes the AI model to misclassify the input. Advai allows data scientists to proactively test AI for issues, and receive actionable insights into how robust their models will be under different scenarios. It also enables decision-makers within organizations to deploy AI models with more confidence.
Cogram lets anyone query data without writing SQL. Users ask questions in plain English, and Cogram generates and runs the matching SQL query. Cogram uses the OpenAI’s Codex model to generate code from natural language.
Inspirient offers the first fully-autonomous, digital-data analytic tool that is able to analyze whole repositories of tabular data without any human intervention. Within minutes, Inspirient’s AI automatically generates presentations summarizing key analytical results and providing business action items. Venture’s first target market is in the survey analytics industry.
IPercept offers industrial companies an automated plug-and-play, secure, hardware and software. The product can be refitted to industrial machines for predictive maintenance, diagnostic and monitoring at machine, sub-system and component levels down to micrometer accuracy, enabling true predictive maintenance without months of training.IPercept focuses on metal cutting, metal forming, iIndustrial robots, hoists, cranes and monorails, general industry and special industry machinery.
Kaedim uses AI software to create digital 3D content from 2D images in a fraction of the time it takes human designers. The venture is initially targeting the video game industry, which has the largest problem of growing production costs for digital 3D assets at scale.
Luffy AI offers a new AI framework software for training neural networks that allow learning and adaptation at the edge, initially focusing on control systems for OEMs and Tier 1 suppliers building autonomous systems. The venture is currently focusing on intelligent thermal actuator controllers, building an adaptive AI flight controller for UAVs, to make performance and learning on real-life edge conditions possible.
QuantPi has built an automated explainability platform that enables every machine learning expert to become an explainable AI (XAI) expert and every non-technical expert to obtain actionable data insights. Their core technology, PiCrystal, consists of proprietary XAI decomposition techniques. It decomposes a large number of XAI algorithms into “building blocks.” After decomposition, PiCrystal creates new XAI algorithms by re-assembling various parts of the XAI algorithms to better address specific explainability tasks. This framework allows companies to combine the advantages of multiple explainability techniques, automatically select the most relevant explanation technique based on specific use-case, and implement trusted and explainable AI models into business processes.
Zeta Motion provides quality control as a service (QCaaS) for manufacturing. Zeta Motion has developed a patent-pending Automated Surface Inspection System (ASIS), which allows rapid deployment and product onboarding, thanks to a proprietary data pipeline and smart AI engine (Neural Capture). ASIS enables manufacturers to improve product quality, maximize throughput, reduce waste, and optimize their workforces.
Lux Aerobot is a space robotics company that specializes in the design, manufacturing and operation of high-altitude platforms for Earth observation. Leveraging its own unique data set, Lux is building tools to support decision-making that ranges from wildfire and coastline monitoring to national security. The company’s first official customer was the Australian Defence Force, which turned to Lux after an unprecedented wildfire season in 2019-2020. Lux now provides a monitoring service that detects and communicates new fires so the Force can combat them before they grow.
Co-founder and CEO Vincent Lachance says there are two aspects to the company: ballooning to take images and AI. At CDL, mentors coached the company to simplify its technology and change its narrative for investors. Lux’s message for investors is now about leveraging the fact that it creates new means to generate data to support its AI ambitions, versus being a ballooning company, Lachance says. “From an investor perspective, CDL is a really good place to be. We’re starting to be good at explaining what we do in AI terms — it has got to be just technical enough, but anyone could understand it,” Lachance said. The honed messaging has not only helped with investors but with communicating the company’s operations to more broad audiences.
The bootstrapped startup recently doubled its team from 10 employees to 20. And while Lux hasn’t raised a round of investment yet, Lachance said he expects contacts from CDL to be main contributors to the company’s first round. He considers the program an exciting place for ambitious founders. “I think CDL is great for that … There are so many smart people that get it. It makes [your venture] feel more alive as a vision.”
Developing a vertical is also about developing a expertize in labelling data in a specific way. A small detail not to forget.
Jean-Francois Connolly • Scientist, CDL-Montreal May 5, 2022 @ 10:11 AM ET