3 of 9

Description of the case study

The Bristol Regional Food Network is building a digital platform to connect local food producers with consumers within a 20-mile radius of Bristol city centre. The platform aims to simplify ordering system and inventory management system to improve efficiency , and support sustainable food systems. Currently, producers manually manage inventory and orders using email, phone calls, and market sales. Customers face challenges in discovering products, managing bulk purchases, and accessing convenient ordering options. The digital marketplace aims to address these issues by offering an e-commerce-like experience tailored to local food systems, while also meeting unique requirements such as managing seasonal inventory automatically, handling the orders and intelligent prediction on the orders to maintain the inventory for future purposes

4 of 9

Tasks

Task 01

System can analyze purchase history predict frequently order items and provide quick re-order options

Note: Utilize synthetic data or can also find online

Task 04

Explainable AI

(XAI) feature can be implemented to provide transparency in predictions

by showing how decisions or recommendations were derived

Note: only require for one main model not for task 01.

Task 03

AI engineers should be able to design and train a new ML model or hybrid model outside the DESD system

and then integrate it with the systems on later stage.

Task 02

Model need to detect the defect and classify the product is fresh and rotten fruits and Vegetable.

According to the quality grade the product and update inventory automatically .

Dataset link:

https://www.kaggle.com/datasets/muhammad0subhan/fruit-and-

vegetable-disease-healthy-vs-rotten

5 of 9

Dataset Description

The Fresh and Rotten/Stale Fruits and Vegetables Classification Dataset is a comprehensive, high-quality image repository designed to support the training and evaluation of image classification models.

The primary objective of this dataset is to facilitate the development of robust computer vision algorithms capable of accurately distinguishing between fresh and spoiled produce. It comprises a diverse range of commonly consumed fruits and vegetables, including apples, oranges, bananas, tomatoes, cucumbers, and carrots. For each category, multiple images are provided, representing both fresh and rotten or stale conditions, thereby ensuring variability in appearance and enhancing model generalization.

6 of 9

Deliverables

Demonstration

First Part : Short Executive Summary 20%

Second Part : Technical and implementation details 20%

Technical Report

Problem Complexity and Technical challenges 20%

Evaluation of findings 20%

Use of Gen AI Declaration 10%

Github Repo

Quality of code and Writing style 10%

Note: details of each deliverables breakdown you can find in the marking criteria in assessment specification document

7 of 9

Major Points to consider

You should consider:
Scalability: Models should handle increasing numbers of products, producers, and customers.
Explainability:

Ensure transparency, customers should understand why they see certain recommendations when they order
Avoid bias (e.g., favouring certain producers in recommendations).
Present forecast charts for producers: for example “High demand expected for tomatoes next week based on seasonal trends and last year’s data.”

Fairness, Accountability and Trust.

What is your strategy for monitoring accuracy over time?
Interactions: If a user overrides the model’s prediction, what are you going to do about that? How does that relate to monitoring performance over time?

8 of 9

Benefit of Integerating AI model in DESD

A single group project addresses the learning outcomes of both the AI and DESD modules, reducing overall assessment workload.

Students develop an AI-based solution deployable as a service, with appropriate interfaces for system-level integration required by DESD.

The integrated project promotes real-time, practical application of AI, bridging theory and implementation.

Enhances student motivation through industry-relevant, hands-on learning and coherent system design experience.

9 of 9

If you are not the part of DESD

You will be assigned to a group accordingly.
It is acceptable for the interfaces to the system to be simpler where this does not compromise the ‘AI’ aspects. During the demo this might mean using Jupyter Notebooks, simple dashboards, and reading/writing files to local storage.

1 of 9