As a participant in the Women in AI (WAI) training program, I embarked on a transformative journey that enhanced my technical skills and broadened my perspective on the immense power of artificial intelligence (AI).
I joined the program after a referral by the Wentors community – an online community dedicated to the growth of women in technology. I was selected to be part of the unique challenge, and I was really excited because was my first internship intensive internship after completing my studies in Data Science and AI at Women Techsters Fellowship.
Through the training, I gained a deep appreciation for how AI can respond to challenges faced by humans. My project involved the use of AI to enhance delivery wait time prediction and reduce the likelihood of delays in e-commerce. I was able to use historic data collected from an e-commerce platform, well anonymized to perform my analysis. Then, I trained a machine learning model on Intel oneAPI to predict the wait time and delay in bringing products to customers.
The project
With the rise of online shopping in the modern world, improving the efficiency and accuracy of goods delivery has become increasingly crucial. This initiative was to serve as a means of solving some of the challenges experienced in the industry, and curb the delay of goods by enhancing the routes used in delivery.
The COVID-19 pandemic has drastically shifted expectations in e-commerce, especially regarding delivery times. My project, therefore, was not just about improving logistics but also about adapting to these changing consumer behaviors and perspectives.
I engaged in comprehensive data collection and preparation, working with a dataset from a Brazilian e-commerce company. This process involved meticulous data scrubbing and anonymization to ensure privacy and data integrity. I focused on developing a dual-aspect model: one for predicting delivery times (regression) and the other for assessing the likelihood of delays (classification).
Implementing advanced machine learning techniques
The project employed advanced machine learning techniques, including XGB, RF, and SVM, for both regression and classification challenges. Then, to refine our predictions, I used a Voting Model, an ensemble approach that combines individual predictions to produce a more accurate consensus forecast.
Then, for the technical setup and performance evaluation, I worked on Environment Configuration. The project utilized both stock and Intel technologies, providing a unique opportunity to compare and contrast their performance. I leveraged Anaconda for setting up the environment, ensuring consistent performance across different technologies.
A key aspect of my role was to monitor and compare the performance of the two technologies, focusing on inference times and prediction accuracy. I meticulously recorded the performance of each model in our testing phase, analyzing metrics like Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) to determine the efficacy of our models.
The project faced some obstacles. One of the primary challenges was the resource-intensive nature of the environment setup, which led us to pivot to the Intel Dev Cloud for more efficient data handling. Handling the large, complex dataset was also significant concern, requiring advanced data management techniques to ensure smooth processing and analysis.
Addressing a specific community challenge
Accurately predicting delivery times and the likelihood of delays is an asset in the African ecosystem, where e-commerce and logistics are rapidly growing industries. The project demonstrated its potential to address specific challenges in logistics and enhance customer satisfaction.
While it employed cutting-edge machine learning techniques, the project identified areas for improvement, such as expanding data collection to include all regions where the model would be used for predictions, integrating it with existing logistics systems and developing mobile apps.
By providing accurate predictions using AI, we can empower businesses and customers. AI will foster a more efficient, resilient, and customer-centric delivery system.
Enhancing the system with NLP
To further enhance the systems predictive capabilities, we can incorporate natural language processing (NLP) techniques. This approach would address the limitation of relying solely on structured data sources, such as historical delivery records, for predictions. By incorporating NLP, the system could extract insights from unstructured data sources, such as customer reviews, social media posts, and online forums, providing a more comprehensive understanding of delivery patterns and potential delays.
The success of this project has far-reaching positive implications for society and various industries.
For society, it would mean reduced frustration and anxiety among customers due to accurate delivery information, improved satisfaction and trust in businesses, promotion of e-commerce adoption, particularly in underserved populations, empowerment of communities, especially during emergencies and unexpected disruptions, and provision of real-time data to customers.
There are many benefits to various industries. For the e-commerce business, there would be enhanced management of inventory levels, reduced shipping costs, and improved customer satisfaction. Logistics companies now have information on optimizing their routes, reducing fuel costs, and improving customer service.
Retail companies can now have more accurate pick-up estimates, and streamline inventory management. Finally, manufacturing businesses can ensure timely delivery of production schedules and materials. For more specific sectors such as healthcare, the technology meant a more efficient management of supply chains and healthcare systems such as the technology used in advance treatment and management of drugs.
Reflections on my AI journey
At the end of my training program, I found myself reflecting on an incredibly enriching and transformative journey. This experience has not only honed my technical skills in artificial intelligence, but has also profoundly deepened my understanding of its impactful applications in real-world scenarios.
This project has allowed me to delve into the complexities of machine learning, engage with advanced technologies, and contribute to a solution that addresses a critical challenge in today's digital-driven society. The opportunity to work with data from a leading e-commerce company and employ sophisticated AI tools has been a highlight of my professional growth.
Beyond the technical achievements, this training has instilled in me a profound appreciation for AI's potential to change industries and communities. The ability to predict information using AI is not just a technological triumph but a significant step towards enhancing customer experience and streamlining logistics in the African ecosystem.
The challenges I faced, from managing large datasets to adapting to different technological environments, have been instrumental in shaping my problem-solving and analytical skills. The experience of overcoming these hurdles has been both challenging and rewarding, providing me with invaluable insights into the practical aspects of AI implementation.
Moreover, the prospect of integrating natural language processing (NLP) to further refine our predictive models is an exciting avenue for future exploration. It highlights the ever-evolving nature of AI and its boundless potential to innovate and improve.
As I move forward in my career, I carry with me not only the technical knowledge and skills I have gained but also a deep sense of responsibility and enthusiasm for using AI to address societal challenges. This training has been more than just a learning experience; it has been a journey of personal and professional transformation, an opportunity to contribute to meaningful change, and a stepping stone into the vast and promising realm of artificial intelligence.
I am eagerly looking forward to continuing my exploration of AI and contributing further to its development for the betterment of communities and industries worldwide.