Projects
Incident Management System Subfeatures (ResponderHQ) April 2025
1. Documents and Messaging
Added documents and messaging features to existing incidents.
2. Interactive Map
Enhanced map functionality used for each region.
3. Collaborative Planning Tool
Collaborative planning tool for resource allocation.
Total subfeatures covered 70% of my work at ResponderHQ.
Image Enhancement Project (ImageCloud) April 2024
One of the projects I worked on at ImageCloud. Worked independently to build an AI photo retouching application.
Technical samples of results:
(I is the input pic, T is the target pic, G is the AI-generated pic)
Application image enhancement process:
In photo retouching, there are usually two types of workflows,
[photo]->[enhanced photo], and
[light version of the photo, neutral version of the photo, dark version of the photo] -> [enhanced photo].
The former is known as a single image, and the latter is known as bracket images. The application facilitates enhancement for both workflows.
The project's model training and deployment architecture was adapated from Google's MLOps level 1 pipeline automation diagram.
For the user facing cloud application, architectural considerations were taken into account to make it cost effective, scalable, maintainable, and extensible. These were addressed by
-
Utilizing lambda functions to kickstart the event driven process
-
Separating each task into containerized images, so that it could be reused.
-
Utilizing asynchronous processing for scalability
-
Utilizing AWS Batch for compute of each containerized image, so that tasks could be independently scaled.
-
Using the AWS orchestrator Step Functions, so that workflows can be easily modified
-
Putting business logic into lambda functions, keeping API focused on routing, and reducing load on it.
-
Using message queue (AWS SQS/SNS) and key value store (Redis) in tandem to provide updates for user experience.
The landing page is a standalone interface for API integration. The final UI/UX will be refined and integrated into the full ImageCloud application by the frontend developer (my manager).
Total project covered 60% of my work at ImageCloud.
ImageCloud Admin App November 2020
A university Final Year Project. Worked in a group of 6 to create an admin app for ImageCloud, a photography business. The admin app was aimed at increasing customer retention and customer engagement with the company features.
My role:
Coded a data mining/machine learning module in Python, which utilizes pattern recognition to predict when a customer has low engagement with a certain feature, then performs user predefined actions to increase engagement.
Tasks:
-
Automated data preprocessing for the machine learning module
-
Implemented a functionality that utilizes a random forest classifier (per business feature) to predict if a customer has low engagement with that feature, based on past data.
-
Implemented a functionality that initiates an extended action, predefined by the admin users, based on the prediction of the random forest classifier.
-
Created and implemented a feedback mechanism so that the random forest classifier can continuously learn from feedback and get more and more accurate at it's predictions.
-
Increased transparency of the random forest classifier.
-
Implemented hyperparameter tuning of the random forest classifier to optimize the model.
-
Implemented analytics based off the random forest classifier model (i.e. factors leading to low engagement), so admins can obtain insights into customer behaviour.
Banking Incident Management System September 2019
A university 3rd year project. Worked in a group of 4 to develop a banking incident management system.
My role:
Coded use cases
-
Create Incident Report
-
Detect Duplicates
-
Assign Priority Rating
-
View List of Incidents
-
Display Incident Report
-
Update Incident Report
-
Implement Risk Management Strategy
-
Perform Root Cause Analysis
-
Show Lessons Learnt
-
Define Roles for Staff
-
Assign Staff to Incident
-
Data Analytics
-
Automated Lock
-
Auto Assign Staff
Implemented Code Review
Website login: user ids are 'b1', 'b2', 'b3', 'b4'; passwords are 'password'.
'b4' is branch manager and thus has admin access.