Founded in 2021 out of questions asked and ideas raised during Miguel Freire‘s and Alexandre Borges‘ masters theses (both researched at INESC-ID and Instituto Superior Técnico), NeuralShift is a new company that develops last-gen artificial intelligence (AI) technologies to build products tailored to companies’ specific automation needs.

From its very start, NeuralShift’s vision has been to integrate computer vision technologies, natural language processing, and sequential decision-making models to bring intelligence to the automation of processes that involve complex tasks (e.g., package sorting, damage identification, and document processing). On the multivariate uses of AI Arlindo Oliveira — a Distinguished Professor at Técnico (Oliveira supervised both Freire’s and Borges’ masters theses), President of INESC and one of the Advisors of NeuralShift — explains, “the application of advanced AI techniques, in computer vision and natural language understanding, has the potential to improve efficiency and customer service, reduce costs and provide data and analytical tools that can be used to redesign processes and services. Last-generation AI techniques will make it possible to offload to machines and algorithms repetitive tasks that are now consuming a significant fraction of available human resources, which can be redirected to higher value-added tasks.“

The best ideas often come from unexpected places. Miguel Freire himself tells the story: “We were both finishing our master thesis in deep reinforcement learning, both advised by Prof. Arlindo Oliveira, when he asked us to solve a problem that was to predict, using computer vision, if a car was damaged or not and where. Our solution ended up going nowhere, but made us realize that these technologies could be more broadly applied to solve novel problems in a number of areas. In particular, a lot of the recent advances in research in deep learning have yet to be applied in industry due to the lack of knowledge about these technologies, about what they can do, and about the type of problems they excel at. We wanted to understand what challenges companies faced and also provide them with a more ample view of the type of problems these deep learning technologies can solve that were hard to solve before. Around April 2021, NeuralShift was created to fulfill this vision.”

AI technologies can have a definitive and lasting impact on business efficient in a way that traditional software simply can’t. “For example, GPT, a transformer based natural language model, enables us to process text in a more holistic way, closer to how humans might do it, in a way that older methods couldn’t previously do with satisfactory results.” Freire adds that “One of our products uses a similar transformer model to process customer interactions. In this way, businesses can provide answers to their customers problems faster.”

Many see the industrial implementation of AI technologies as inevitable, and Freire agrees. “In the next years it is expected to have AI based-products in every industry, so for a company to be competitive it must use AI. We see ourselves helping businesses shift to this new paradigm and we also want to advocate for the responsible use of AI. We also want to guarantee that our solutions follow a number of principles, including explainability and fairness.” Freire is clear: “we should not only try to model the data as is, so that the model is effective at solving problems, but also as the world should be, so that it helps bring a fairer future. We’re currently a member of the Responsible AI Consortium, which includes other Portuguese startups, research centers and companies, all working towards the goal of AI technologies that are fair, transparent, privacy-preserving, and energy-sustainable.”

NeuralShift is one of several spin-offs that have been initiated at INESC-ID by researchers and ex-students over recent years. To have a look at some of these other companies please check out our spin-offs webpage.