Client Need:
Selling thousands of products a day means you have substantial amounts of paperwork, shipments, returns, and storage allocation that need to be supervised. At the time, our client was handling all of this manually which brought about numerous issues; losing track of products, uncertainty about a shipment’s payment status, inability to keep track of returns, lack of storage management, among others. They wanted a digital platform that would consolidate and automate these internal processes, cutting down on workload and greatly reducing the risk of user error.

Research:
Our objective during the research phase was to gain a comprehensive understanding of our client’s operations. This primarily consisted of collecting data from various sources such as online sales, courier services, storage facilities, bank statements, and returns. The main challenge here was managing diverse data from multiple countries and bank accounts, so we organized additional discussions with the client to ensure that our interpretations were accurate, and that all possible cases were accounted for.

Development:
Upon finalizing the project scope, our team opted to utilize a technology stack composed of Node.js, React, MySQL and Python to develop a custom dashboard solution for our client. The objective was to create a centralized platform enabling the client to:
• Import data from various sources, including couriers, banks, and online shops, across multiple countries.
• Manage custom files related to returns.
• Track and manage packages using a range of filters.
• Oversee multiple warehouses.
• Manage user access and log their interactions within the dashboard.

Working collaboratively in a team of three, we successfully delivered a tailored solution to meet our client's requirements. The process involved uploading bank statements and delivery orders via Excel and CSV files, which were then processed using Python before being sent to our Node.js server to be stored in the MySQL database. Utilizing MySQL queries, we were able to cross-reference data between files, identifying unpaid orders and orders awaiting shipment. Similar queries were utilized to track real-time inventory levels for effective stock management.

Communication:
The extensive variance in the inputs meant that communication in this project was paramount. Weekly meetings were held to ensure that all this information was communicated, and for our client to provide feedback on the UI, which would then be subsequently addressed.

Deployment & Support:
Using Nginx, we deployed this app to its own Linux server where we can monitor the status of the app. Since this software went live, our client processed over a hundred thousand records, automated their warehouse management system, and saved considerable amounts of resources by using our solution. After deployment, inputs were changed numerous times but we delivered updates quickly with little downtime so that our client could continue using the software without disruption or any losses of data.

Client Need:
Expertize of EMAS lies in the quality of their predictive models. To present the results of their work to their customers, they needed a software solution that would collect the data required, send it to the scripts that will run predictions, and lastly, present it to the customer.

Solution:
We built an application that can explain the results of the model to the Account Managers with zero technical knowledge. It was done using Dotnet at the core of this application, Next.js to present the results, Python for data processing and running of ML models and PostgreSQL for the database.

The Client:
BEX Express is a courier company providing domestic and international parcel delivery, package tracking, and courier scheduling services. They offer multiple package drop-off and pick-up locations across Serbia.

Client Need:
When users of their platform send them shipping information, they make all kinds of mistakes. Some letters are missing, or they have extra letters; writing numbers using words; misspelling; etc. Because of these mistakes, their system is unable to find these locations, and the error must be fixed by hand. About 40% of the daily enquiries well into this category.

Solution:
By designing an algorithm that would parse city names, street names, and other information they provide, we were able to find 95% of the cities and municipalities in their database and 68% of all the streets.

Software:
The requests are made to the POST method of Python Flask. The json contains information that the user provided. The json is then parsed and sent to separate Python scripts to find the city and municipality and the second script to find the street name. These scripts are generating all possible combinations, using provided input, that might exist in the database. Each combination is then sent to the SQL script, trying to find it in the DB. In the API response, we either send ID of the found locations or some of the best guesses that will help customer service find the missing location.