While traditional restaurants are struggling to survive in the midst of Covid-19, new-age cloud kitchens are swiftly taking their charm and customers to the online world. Rebel Foods started as a brick-and-mortar restaurant and then transitioned into the cloud kitchen business. From single brand cloud kitchens, the company built multi-brand and multi-channel cloud kitchens over time.The word “Cloud” in their business model stands as the cornerstone of Rebel Foods. The born-digital company has built its entire revenue and operational structure on cloud technology.
“For every challenge which we have faced as we embarked from traditional retail to cloud kitchen, we have always considered a technology solution. We have been using AI for monitoring and auditing our kitchens and also tracking daily activities. Our team has implemented AI to ensure health and safety while retaining the customer,” Soumyadeep Barman, Co-founder & Chief Product Officer, Rebel Foods told ETCIO.
The company has invested in Microsoft’s AI technologies to monitor/audit its kitchens across hygiene and SOP adherence day in and day out. They have adopted an AI-based video analysis solution, Jarvis, built on Microsoft Azure, and developed by a Microsoft partner, Staqu, that constantly monitors kitchens for health and safety violations.
The unique part of the company’s approach towards leveraging AI is to retain customers. Amid COVID-19, Barman realised that customers are willing to spend on food orders but they need an assurance of hygiene, sanitation, and health safety apart from the quality which the company already used to serve.
“In the initial months of the lockdown, we saw a drop in business, given the sudden scare of the virus in patrons. Today customers are not satisfied with information such as the body temperature of delivery boys, etc. They need to know the entire process from cooking to delivery, what is going behind the scene with their order. Therefore, we planned to implement AI which can help us in visualising and monitoring the kitchen process. It also helps in audits as well as gaining customer’s trust,” Barman reiterated.
Barman also explained that the AI solution called Jarvis, uses the video feeds from security cameras inside the kitchens to check whether the staff members are wearing caps, gloves, and face masks and whether the floor is clean, and, of course, if food is being cooked properly. It flags lapses in real-time enabling kitchen managers to act immediately and not retrospectively.
Jarvis is built on Azure and after evaluating the success of the initial deployment, Barman has already conceptualised nearly two dozen additional use cases with AI. Microsoft.
Starting off with the AI use case
AI has been built under the automation umbrella largely. At Rebel Foods, the idea is to use AI to instill faith and trust in customers by having total transparency. The video feed captured, analysed, and approved by the AI solution is streamed to customers. Planting trust and transparency as the core of their business.
The company has also built an AI system to scan a particular product from Size, Weight, Appearance, and Temperature metrics and decide whether it can be served to a customer or not. The system is like a gatekeeper at the kitchen and is trained for all products in the company.
“Though finding the solution to restore customer trust and what needs to be done was important and key, deploying at kitchens and making sure the models are running perfectly fine, was more challenging. Training the data models with 20 odd use cases of hygiene and SOP adherence, with every individual and kitchen construct being different, was an uphill task,” Barman emphasised.
Barman and his tech team started with one use case at a time, worked on them, measured them against manual audits side by side to understand accuracy, and then took them live. From a business perspective, it was to give control back to customers by increasing transparency and hence increase trust as a result, and from the tech side, it was to have real-time audits/alerts and streaming for total visibility.
Cost challenge is the real struggle
One of the biggest challenges that Barman and his tech team were facing was to have real-time video analytics, alerts, and streaming on the cloud without increasing much infrastructure costs at the kitchens. His team evaluated several technology solution providers and then chose Microsoft to continue with.
“We started to implement a scalable and flexible solution powered by Microsoft Azure. In this project, the platform developed by Staqu didn’t push us to replace our old cameras as our app, which runs on Azure, could connect to existing cameras. This process was made seamless with Azure Kubernetes Services, which scale the product by adding new AI models and features remotely,” Barman highlighted.
Strategy to tackle AI Bias
Rebel Foods is on a tech transformation journey that still requires a lot of work at its end. Primarily, to reduce AI bias, the company focuses on the following parameters- firstly, to define the problem well with clear success metrics to avoid scattering of the situation and expectations. Secondly, they focus on designing and understanding the training data aligned to the success metric during the definition stage, keeping a completely open mind for all use cases. Lastly, the team develops and delivers the solution with a mindset of listening to feedback and retraining the models continuously.
“Most critical is to have multiple open channels for feedback (integrated/open forums) to keep improving every moment. Our partner Staqu uses Azure, which is bound by Microsoft’s Responsible AI policies. The policies rest on the principles of fairness, reliability and safety, inclusiveness, privacy and security, accountability, and transparency,” Barman added.
