How Crater Labs’ Saved Millions With a New Scheduling System


Scheduling should be easy — it shouldn’t cost you your profit margins. Learn how Crater Labs used machine learning models to save their clients millions.


Background


Crater Labs is a Canadian machine learning (ML) research lab that helps businesses interested in adopting AI achieve their biggest tech problems. By collaborating closely and emphasizing out-of-the-box thinking, they’ve been able to secure massive wins for their clients, including life-saving radiation reduction and life-changing software bias elimination.


Read on to learn how Crater Labs helped a multinational engineering firm save millions by simply improving their scheduling system.


Challenge


Most large-scale construction projects are government-funded initiatives like nuclear construction projects or transit system improvements. With high stakes comes better money, but the bigger the project, the more employees you have working in sensitive locations — sometimes for over a year.


Ensuring that these locations remain secure and accessible only to authorized personnel is a significant challenge, but not quite as difficult as scheduling a locked-down project in an efficient, accurate, and, most importantly, secure way for hundreds of employees.


What’s more is their research found that nearly 2.5% of tradees couldn’t do their jobs week over week due to materials or equipment being unavailable, adding millions of unnecessary waiting time to the project’s budget.


Solution


Crater Labs is first and foremost a research lab. Undertaking this project would require a lot of data, a fair amount of time, and some regular human input.


It also required three different machine learning models to address the client’s three major issues:

  1. How long workers were in secure areas, therefore how long their potential radiation exposure was

  2. How much overtime was required to complete the project

  3. How to keep up with the ever-growing possible compliance violations

Because all machine learning models need some form of human interaction and data input in order to begin training, Crater Labs tapped into the client’s existing scheduling database and fed it all into a cluster-based local factor algorithm (an unsupervised model).


The resulting information was then fed to a supervised machine learning model to identify potential overages, and ultimately evolved into a semi-supervised model that learned about every authorized access request and possible violation — without the ongoing need to create more rules manually.


Conclusion


Using the three problems to then create associated ML models allowed Crater Labs to take a slow, intentional approach to their client’s AI solution. It also allowed the client to feel apart of the process, which is vital for any AI service provider.


At the end of the project, the client walked away with a completely personalized long-short-term-memory (LSTM) neural model to account for their irregular distributions in real-time.


But more importantly, at the time of delivery, the model predicted future radiation exposure with 99.1% accuracy and flagged work completion issues with 92% accuracy, ultimately saving the client millions in project management costs.


Contact


To end your project inefficiencies and get your team back on-track, reach out to us directly to learn how affiliates just like Crater Labs can get your AI journey started (or improved!) today, or read more case studies here.

18 views0 comments

Recent Posts

See All