Eliminating volatility in your supply chain is impossible, but managing it is not. Trusted by top brands, Kinaxis (TSX:KXS) is the leading provider of cloud-based, software-as-a-service (SaaS) solutions that give people the confidence to know they are making the best supply chain planning decisions to maximize business performance. We solve complex business problems in easy-to-understand ways by combining human and machine intelligence to plan for any future, monitor risks and opportunities and respond at the pace of change. With the support of our community of supply chain experts and using our unique concurrent planning technique and single integrated planning platform, customers can realize higher revenue, lower costs and fewer risks.
Kinaxis RapidResponse is used for concurrent planning, instantly and continuously balancing end-to-end supply chain networks. Our vision is to optimize supply chains across industries, in order for products to reach consumers faster and reduce unnecessary waste. While creating superior client and supplier experiences, eliminating wasted time, resources, and talent benefits the environment and communities across the globe.
World is operating faster than ever, and larger scale systems such as supply chains require more than just spreadsheets. There is so much more data today. Historically Retailers collected data but did not know the best way to leverage them. Now with AI, they can turn data into a competitive advantage with more accurate forecasts, react faster to changes in the market, and play out counterfactuals or future scenarios.
If I’m the owner of a supply chain, I usually have data specific to my operations and my company that reflects the reality of factories, inventory, distribution centres, raw materials and other components. Now bringing this specific data together with the abundance of public datasets such as weather data or financial indices, Kinaxis is able to integrate multiple data sources to produce a rich and hybrid picture for forecasting.
In all cases (students and ft), every new team member is assigned a personal mentor --someone who can answer all their questions. There is also a separate ongoing/self-serve mentorship program, where people can request mentorship pairings.
There is a culture of helping each other out: you will be encouraged to ask questions and participate in forums for knowledge sharing. For example, you grow by attending research meetings where a data scientist may share her findings with the extended cross-functional team. Opportunities are ample to learn outside of engineering about the general business through marketing or product team meetings. These conversations are concentrated on problems about the wider industry or customer demographic.
In the ML team specifically, we encourage everyone to present technical concepts to improve how they communicate technical information to a business audience.
Every quarter, we hire several interns in ML specifically, to help with applied research for a real customer problem. We never ask students to come in and fix bugs on legacy systems.
Out of a pool of projects, we work to find the best fit based on the intern's experience and personal interest. Projects range from backend-oriented heavily touching on statistics and math theory, to more frontend/user driven, such as building informative data visualizations. You are paired with a mentor at the beginning, and usually wrap up with an executive presentation. It’s common to see interns co-authoring a patent, writing published papers, and speaking at events and workshops. We also welcome return interns, who are interested in exploring different parts of ML.
There are also two other types of internships: interns that contribute to software on our scrum teams working on product, and academic internships with a research focus.
When working on applied research, it’s important to prove your results. Data science experiments usually take place within notebooks and visualizations, but it is all rooted in the scientific method. Start with hypotheses, experiment, measure, and demonstrate the value and validity of your code on a customer dataset.
To sum it up in three words, the attribute we look for is “creative problem solving”. Experience with some ML libraries is good but there is no hard requirement to have used a specific set; rather, we are interested in what you did with the tools and the complexity of the problem you chose to solve previously.
We recommend describing the work that you did using less common datasets rather than those typically used as part of introductory ML courseware. To stand out, selectubg an unusual or interesting dataset, then doing something nifty and out-of-the-box, perhaps under the guidance of a professor, that goes beyond the tutorial level and takes on a real and new challenge. Supply chain management problems are almost always the latter kind although your experience need not be in supply chain specifically.
It’s helpful to compare ML problems in the supply chain to a canonical task like image recognition. In a normal classification task, the definition of classes does not change, so models tend to improve by viewing new examples, and transfer-learning is possible via adaptation of trained models.
In our field, we are working with time series and temporal dependencies; whatever is true just a month ago could be false today because features are affected by new events that change customer behaviour. Instead of data that is largely static, supply chain activities generate stochasticity and data-drift, which in turn becomes extremely hard to predict. Moreover, we need to mix and match data that is not at the same granular level. For example, weather data could be at the location level, sentiment data could be at brand level, and shipment data at the distribution level. The distinct scales in features make them difficult to combine. Lastly, everything needs to be done at scale, and systems need to answer business questions to plan in real-time. Planners demand answers in minutes from a production model.
But these are also the reasons that make the problem special and interesting to tackle.
We value diversity and advertise in different channels with the intention of reaching out to various (student and professional) organizations to increase the visibility of our job opportunities. Beyond distributing the hiring information, we like to interact in the community by regularly hosting workshops at universities and mentoring hackathons for under-represented populations.
Kinaxis set out to create a Diversity, Equity and Inclusion (DEI) program that makes us stronger and that reflects who we want to be as a company. As we grow, we are becoming more intentional about programs and initiatives that support an inclusive and diverse environment.
Having a better understanding of our workforce and the needs of all employees allows us to set goals and take actions that benefit everyone. We understand that together, we are stronger and that supporting all colleagues is a top priority. Diversity, equity and inclusion is everyone's responsibility so we are working towards our goals in the way we know best, together.
Recruitment plays an important role in DEI. Our DEI Recruitment efforts allow us to recognize and help eliminate biases within hiring practices which may be related to a candidate’s age, race, gender, religion, sexual orientation, and other personal characteristics that are unrelated to their job performance.
For students: After a rigorous selection process, we schedule a 30min interview for selected candidates. Sometimes, there is a second follow-up interview and a data science technical screen. We use tools like Hackerrank to host a collaborative coding session with the candidates to understand basic Python abilities, and follow-up with a ML Rapid Fire on the conceptual level.
For full-time: Automated python coding test evaluating style and correctness; phone screen (hiring manager); first set of interview (HR, another longer interview with hiring manager); on-site interviews (key-members of the specific team and its director). Mix behaviour and technical.
In general, previous experience in ML or data science internships is a plus. The application of the relevant technology to a problem is great, which does not have to be in our domain but ideally with a business application. This can take the form of hackathons and personal projects.
We have teams focused on the areas of ML platform, data workflow and infrastructure, research and development, and productization. Within teams there are usually breakdowns by specific projects and how AI/ML is applied. First, the team conducts applied research with some data to figure out generic models to pursue. Then, we undergo standard software development life cycles to create the final system and product.
