I had worked on this project, whilst at Optimtech Technologies for a large local bank, in Pakistan. The application was developed on Spring and made use of the Oracle Database. I was responsible for extending the backend webservices and making changes to the document management and document state machines, which were coded in Java. I was also the liason for this project, hence I was responsible for communicating with the client and relaying the requirements to the front-end team as well.

This was a private project, which I had undertaken as a contractual job, with a small local company. They already had ERP Next deployed at their location for managing client accounts and were making use of QuickBooks desktop to manage financial reporting and billing for other companies. I was responsible for developing a mechanism to sync the data entered into QuickBooks, with the data already present in the ERP. For this I made use of the Quickbooks Desktop API, to pull out data into a local instance of MySQL. Once the data was in the database, I developed a syncing mechanism for the database and the ERP, through a Python drop service.

I had recently helped a startup develop an extensible and secure development framework for their IoT offerings. In this project, I was responsible for developing a mechanism to share data between the end-IoT devices, with a centralized cloud, in a secure manner. I was responsible for coding the data collection mechanisms and the data transfer protocols within the devices. The central server was made to serve as the Edge Device and the single point of contact with the cloud. I was also responsible for coming up with a containerized mechanism to remotely deploy code on the edge and directly connected devices, without the need to be physically present at the location.

As part of my final year project, I wrote a research paper, which is pending publication. This was for a proof-of-concept for predicting natural disasters, so preventative action can be taken before-hand. I made use of large-scale satellite imagery to identify regions susceptible to a particular natural disaster and correlated it with other sources of data to make my predictions. Due to the large size of the satellite images, certain optimization and multiprocessing techniques were added, so as to speed up the processing time. For image processing I made use of machine learning and for predictive modelling a regression technique was used.

As part of another research project, which has already been published, I was part of a team that developed a mechanism to train neural networks in a distributed manner. This team was comprised of PhDs and I got to learn a lot from them. I was responsible for coding the neural network in PyTorch and coming up with a mechanism to deploy the training part of the network on a hadoop cluster. I was also responsible for the setup and configuration of this cluster and ensuring that the accuracy of the training mechanism was not compromised due to the distribution.