DigitalTransformationTalk: AI and Data-Powered Manufacturing

On 28 November, Digital Transformation host Tim Long, Global Head of Manufacturing, Snowflake  was joined by Ben Pecson, Data & Software Engineer, Harkins Builders; Quinn Killough, Computer Vision Specialist, LandingAI; and Tim Long, Global Industry GTM Lead, Manufacturing, Snowflake.

 

The Snowflake platform

Manufacturers are now looking for different approaches and strategies. It’s no longer about having the best quality at the cheapest price but also about building relationships with customers and engaging them.

 

Continuous development is replaced by transformation enabled by a digital platform where all data, both structured and unstructured, can be integrated. The Snowflake platform can also provide manufacturers with both an upstream and a downstream view including data from customers, suppliers and other partners, as well as a marketplace where businesses can tap into economic statistic about the market. Snowflake, in essence, is the AI data cloud that offers data lake and data engineering capabilities and – what may be especially  appealing to manufacturers – industry partners that have already successfully delivered on manufacturing use cases leveraging the strengths of the Snowflake platform. 

 

Use cases so far range from equipment repair and maintenance and energy optimisation to anomaly and defect detection to demand forecasting and supply risk management. The platform provides great data and model governance through capabilities known as Snowflake horizon and model registry, respectively, as well as two different toolkits – one for GenAI and one for ML. It can take a document several dozen pages long – for example, from SEC filings –  and answer a very precise question about supply chain risk

Landing AI: Gaining value from GenAI 

Landing AI helps companies in the semiconductor, automotive and medical sectors to build, deploy and scale visual AI solutions based on Large Multimodel Models (LMMs). Its flagship product is LandingLens. Use cases fall into several different categories ranging from defect detection and assembly verification to robotic guidance to OCR. The main challenges of adoption that businesses should be mindful of are data quality and availability, development time and ROI and balancing cloud and edge computing – the latter is key to be able to operate at speed for real-time inspections. Sometimes a business decides to piece together a selection of different tools with a data scientist leveraging open source solutions, which later turns out to be hard to scale and share.

 

LandingLens has three major modules – one labelling the images, the second training the model with the click of a button and evaluating the model before it’s taken to deployment. At the end of the cycle, production data is fed back to the development environment. LandingAi is able to run 100% within Snowflake. A manufacturer managed to automate a manual inspection process with LandingLens and outperform highly skilled manual inspectors thanks to implementing the solution. 

 

Smoothing the transition from planning to construction

Turnover is a stage where the marketing team of the planning business advertises the project through fact sheets and resumes. Harkins Builders uses GenAI on the Snowflake platform to create the turnover narrative for the company leveraging the data that sits in the company’s CRM system. Harkin Builders estimators can view all the data that’s stored on Snowflake about that particular project and generate prompts for Snowflake Cortex. The prompts generated this way are then sent to Snowflake, which returns a coherent, grammatically immaculate turnover narrative written in the desired tone – which can be further improved via additional prompts. Harkins Builders have further plans to use the Snowflake platform for financial forecasting, anomaly detection and preventing safety incidents.

The panel’s advice

• Let your non-tech staff know that you as a developer have a receptive ear for ideas about AI.
• Always follow several different paths of using AI and learn from initiatives that eventually didn’t come to fruition.
• Using ChatGPT may require too much overhead and maintenance and tailoring compared to a platform like Snowflake.