Hashan Narangodage

University of Alberta

Certificate, Software Product Management

January 1, 2020 – January 1, 2020

University of Alberta

Certificate, Software Processes and Agile Practices

January 1, 2020 – January 1, 2020

CIMA

Advanced Diploma in Management Accounting

November 1, 2013 – Present

Curtin University

Bachelor’s Degree, Computer Systems Engineering

January 1, 2011 – January 1, 2015

ASAPP

Product Manager (Rotation) - Intents and Coaching Insights

May 1, 2022 – August 1, 2022

Hybrid

ASAPP

AI Research Scientist / Research Product Manager

February 1, 2022 – Present

Hybrid

ASAPP

Machine Learning Research Engineer

October 1, 2018 – February 1, 2022

Hybrid

WSO2

Solutions Engineer

August 1, 2017 – October 1, 2018

Colombo, Sri Lanka

LSEG Technology

Software Engineer (Machine Learning)

November 1, 2013 – July 1, 2017

Malabe, Western Province, Sri Lanka · On-site

Tekkit

Director Of Technology

March 1, 2012 – February 1, 2015

Colombo, Sri Lanka

Breast Cancer Detection and Classification

October 1, 2016 – Present

Trying to develop a full pipeline that can detect and classify breast cancer using mammogram analysis. While we are still researching on this topic, the expected final outcome is a system that takes breast mammograms as inputs and alerts if the mammogram shows abnormality and also whether it’s benign or malignant.

Domain Specific Language Development (DSL)

December 1, 2015 – December 1, 2016

This DSL is at an abstraction level close to the business specification, and uses deterministic inference to allow system definitions that are as much as 500 times more compact than the corresponding code. This revolutionizes the system engineering process by radically reducing development time, improving system quality and robustness, and reducing management overhead. The toolkit we developed generates a complete running system including APIs and user interfaces from the system definition written in our DSL. This was a substantial project which is now being prepared for production use.

Towards measurement for source code quality

January 1, 2015 – December 1, 2015

The goal of this project was to develop a statistical model of source code. A model that goes beyond syntax, and by learning from a large collection of “well written” code, attempt to model code semantics at a rudimentary level. I approached this task by building a model that extracts patterns present in the code it is trained on, treating patterns in code as a proxy for semantics. Once trained, the model would give higher scores to unseen code that has similar patterns. This research was based on finding a good decomposition of the code’s Abstract Syntax Tree (AST) - a decomposition that does not loose too much information about the long distance relationships present in code. Our model encodes only the structural information of the AST, and not the surface tokens of the program code, and once trained gives lower scores to code similar to the training set. We tested this model across two different programming languages (python and Java), on 7 different projects. Used python source code belongs to two different domains (scientific computing, web) and, used Java source code belongs to 3 different domains (database, build tools, search engine). In every case, given previously unseen code, our model was able to identify both the project and the domain with 100% accuracy.

Step Climbing Robot

September 1, 2013 – Present

We built this robot for a competition at SLIIT. It is capable of climbing steps that are 5 inch tall and can follow a line to get to that object that it is suppose to pick. We build a hand into this robot to pick up the object from a specified destination and drop it at a specified destination. There were few junctions on the way as well.

GPRS Activated Door Answering System

September 1, 2013 – Present

It's basically a door answering system with the capability to call the owner of the house if they are not at home. This system can be activated to work with both the GPRS technology and through the landline. This system will be a very important thing to the society with the busy schedules, because lets say if somebody come to your house to give an urgent message and you are not at home, and that person doesn't have your mobile phone number. What happens then is it will be too late when you hear the message. But with this system in place that will not happen, because if you are not at home system will automatically give you a call connecting that person with you through the in built speaker and the microphone, that way you will get the message then and there. So we think it's an essential system for the society nowadays.

Train Platform Direction System using an FPGA

March 1, 2013 – Present

The system will automatically detect whether a train is coming using weight sensors and it will automatically close the gates in the railway crossing that way we can avoid accidents in the railway crossings. This automated gate system is important because if we assign a person to do that job it'll be a waste of human labor because he won't have much work most of the time in the day. And by using this system we can avoid human errors as well. Another importance in this system is that when it comes to a train station it will show the train driver which platforms are free for him to go to. We can reduce human errors in that specific task as well by automating that. We were trying to automate the already available systems to increase the accuracy of those tasks, that's the main reason we used a FPGA to do this task as well. Because we can do high speed computing using an FPGA so the system will be very accurate.