Indian-based AI platform company CyborgIntell has put its Africa ambitions into action having opened an office in South Africa. Although the company has not specified plans as far as expansion is concerned, it is confident that its low-risk, solution-focused value proposition will help businesses in Africa.
CyborgIntell was founded in 2018 in Bengaluru, India, by Suman Singh, Amit Kumar and Mohammed Nawas. The CyborgIntell platform addresses the key challenges companies face in the data science/machine learning lifecycle – from data selection and modelling, operationalising AI, to managing risk and governance.
According to the company, the product was market ready in the second half of 2019 and while it has established partnerships and clients in South Africa since then, plans to enter the South African market directly were put on hold in 2020 due to the COVID-19. These are now being actioned.
Headed by Managing Director Bryan McLachlan, who has over 30 years’ experience in financial services, CyborgIntell Africa will work closely with financial institutions and other enterprises to help them rapidly develop, deploy, and operationalise AI applications at scale.
Specifically CyborgIntell aims to assist businesses that continue to struggle with challenges in the data science/ Machine Learning (DS/ML) lifecycle.
CyborgIntell said the lifecycle of building and deploying AI/ML usually includes:
• Data preparation: pulling large volumes of data together (usually from multiple sources) into a single format. Deriving, identifying, and creating hundreds of additional new fields of data from the existing data, designing the analytical solution.
• Model building: ingesting the prepared data and, using multiple statistical techniques, developing models that can make predictions or classifications.
• Deployment: taking complex models and enabling them to work in an operational environment.
• Model Risk Management and governance
• Effective decision making, often with a human interface or intelligence
“Each of these is a separate but integrated module in the CyborgIntell platform,” sad McLachlan.
And the main challenges that give users headaches are listed as:
• Talent: Traditional DS/ML projects need a cross-functional team of subject matter experts, data scientists, DevOps engineers and developers– which is expensive and complex to build, retain and manage.
• Time: Time to deploy can be long (six months or longer) due to multi-functional complexity and the manual work involved in developing, deploying and operationalising ML models. A lot can change in six months, which can affect the performance of ML models. In addition, there can be a significant opportunity cost due to time delays.
• Governance: Organisations need a comprehensive control framework to regulate AI model development, deployment, monitoring and management of models.
• Trust: Explainability and interpretability of ML models is important to gain trust and confidence. Companies should know exactly how a ML model got to a recommendation or decision. ML models are at risk of bias and overfitting, resulting in AI failure.
• Risk: Financial institutions need a robust model risk management framework to assess and mitigate risks associated with AI and risks of AI failure.
McLachlan said: “AI is a powerful and transformative technology, yet many companies across the world find it difficult to unlock its full potential. More than a third (36%) of organisations take more than 90 days to deploy data science machine learning (ML) projects, while the failure rate of such initiatives is estimated to be 85% across industries.
Why is the failure rate so high?
McLachlan explained: “For example, the cost of a multidisciplinary team working on AI project for months can be high, making it difficult to achieve a good return on investment. Trust can be a big issue. In a credit application example, a company should be able to explain in plain language why an application for credit is declined. This is often not the case. It is also important to effectively monitor model performance for deterioration, which can be expensive if it leads to poor decisions. If a model fails it can become costly for the company if it can’t be retrained and redeployed in a few hours rather than in months.”
CyborgIntell said it has created a platform that enables enterprises to accelerate adoption of AI and ML by operationalising sophisticated ML models for effective data-driven decision-making, within two to four weeks, with transparency, explainability, governance and risk management.
“Our vision is to help African organisations extract the best returns from their investments in data science, AI and ML by automating the data science and machine learning lifecycle,” McLachlan reiterated.
CyborgIntell reduces the time required to develop accurate, production-ready models to just a few hours without writing any code. The scalable AI platform can address a variety of use cases for every enterprise in various industries. Furthermore, it enables organisations to interpret, explain, and trust ML models. It understands, mitigates bias, and continuously improves performance, ensuring AI adoption.
CyborgIntell’s next-generation approach offers a one-stop, zero-code solution for rapidly developing, deploying and operationalising AI applications at scale. This approach slices the time to deploy AI, while helping to reduce risks and enhance ROI.
McLachlan said: “We are excited to be investing in Africa with a view to democratising AI and helping organisations unleash their full power. If a company has data, and a business problem, we can probably help them. Part of the opportunity in Africa for CyborgIntell is that we address the skills challenge. CyborgIntell’s platform has an “expert experimentation mode” for advanced data scientists, as well as an “auto mode” that lets business users build and deploy machine learning models. It is feasible for companies early in their digital journey to leapfrog to next generation AI technology.”