Author
Noah Xiao
At jahan.ai, we specialise in using AI/machine learning to maximise value for businesses in retail, CPG, and manufacturing.
Today, we're excited to share a pivotal decision that is setting our demand forecasting product apart: Using the Julia programming language.
The New Scientific Language in Town
Julia is an exciting and relatively new open-source scientific computing language. It is fast, dynamic, and easy to use, making it a perfect fit for our demanding task of forecasting over 20 million time-series for some of our clients. We believe Julia is instrumental in delivering seamless, accurate large-scale forecasting.
Embracing Best-in-Class Algorithms
Our journey began with our commitment to excellence, aiming to offer not just solutions, but the finest in the field. That means there is an essential need to incorporate top-tier forecasting algorithms like XGBoost, LightGBM, EvoTrees, and Flux — all stars in their respective arenas of machine learning and deep learning. Julia, known for its high performance, provides excellent compatibility and support for these algorithms. Its ability to handle large and complex computations and mathematical processes seamlessly allows us to provide our clients with the most accurate, refined, and effective forecasting models in the market.
Performance, Performance, Performance
Handling enormous data for retail giants, including forecasts for over 20 million series in near real-time, demands a robust, efficient, and cost-effective language. Julia fits this bill perfectly with its minimal latency and high-performance capabilities. It allows us to scale effortlessly, processing vast datasets quickly thanks to efficient memory usage and parallel processing. This speed is crucial in an industry where trends shift rapidly. Julia's just-in-time (JIT) compilation and performance, rivaling that of C, enables us to update forecasts multiple times a day. Its unique blend of speed and interpretability, without sacrificing power, helps our clients to react proactively to the ever-changing market demands.
Accelerating Delivery Timelines
Our clients (retailers, brands, and manufacturers) operate in high-velocity environments, and making timely decisions is crucial. Therefore, our ability to build and deliver solutions quickly is paramount. Julia’s ease of use, combined with its expressive syntax, has significantly condensed our development timelines. Our team can quickly iterate over multiple models, fine-tuning applications at a remarkable pace. This agility helps us deliver prompt services to our clients, helping them stay ahead of the curve.
Born for Operations Research
Another strength of Julia is operations research. Beyond the well-known realms of predictive and generative AI/ML, we see mathematical optimization as a crucial component of data science. Optimisation of various operations is crucial in retail, CPG, and manufacturing, from promotional and pricing optimisation, to supply chain management and replenishment, to inventory optimisation and workforce scheduling. Operations research (OR) is at the heart of these tasks. Julia stands out prominently in this domain. With packages like JuMP (Julia for Mathematical Programming), we can formulate and solve various optimisation problems with ease.
With Julia's efficient numerical abilities and a comprehensive ecosystem of operations research libraries, we efficiently tackle linear, mixed-integer, and complex programming challenges, ensuring optimal strategies that deliver greater value to our business and customers.
The Growing Ecosystem and Community
The growing Julia community is a testament to its rising significance. Top organisations like Google, Amazon, and NASA have adopted it for various use cases, recognising its potential in large-scale applications. Leading universities such as MIT, Stanford, and UC Berkeley are teaching Julia, preparing the next generation of data scientists and engineers with this powerful tool.
While the Julia ecosystem is still young and evolving, with a long journey ahead, jahan.ai is dedicated to actively contributing to its growth. We're focused on enhancing the ecosystem through more open-source opportunities and fostering a collaborative environment for innovation
Conclusion
We, at jahan.ai, take pride in selecting the right tool for the right job. Adding Julia to our tech stack is more than a decision; it's a commitment to excellence, efficiency, and evolution. We are not just adopting another programming language; we are adopting an innovative culture that pushes the boundaries of what's possible in retail demand forecasting.
If you share our passion for what we do and are excited about contributing to the Julia ecosystem, join us on this exciting journey forward!
At jahan.ai, we build end-to-end AI twins for retail and supply chain businesses. We take the concept of a digital twin, a virtual replica of business processes, further with advanced AI, which not only mirrors these processes but also automatically optimises them. In this way, we support businesses in driving efficiency, going beyond their potential.