Tech
HackFrost 3.0 Hackathon - 1st Place Winner
HackFrost 3.0 was a hackathon targetted at tackling emissions from industrial transportation. Our
team, Viridis,
designed a DIY kit for retrofitting long-haul trucks with solar-powered Auxilliary Power Units
(APUs) - a solution
that is near-immediately deployable, scalable and with the potential to annually offset 1.5x
NYC's annual emissions
every year if deployed across North America. Over 150 people participated at the event.
Mysa's Group Photo - 2019
Mysa's group photo in 2019 (with me ecstatically pointing at the Mysa logo 😂)
Googleplex - Google CodeU Retreat (2018)
Google CodeU is a 12 week invitation-only virtual program, composed of 150 high potential
undergraduates
pursuing Computer Science in North America. At the end of the program we were hosted over for a
3-day retreat
at Google's Headquarters in Mountain View :)
Bounce Hackathon - 2023
The 2023 Bounce Hackathon was targetted at addressing employment retention issues in the health
sector
within NL. As part of this Hackathon, I developed a rough simulation of the impact of employee
churnout
on the workload of remaining team members, and how the current rate of churnout relative to
employment
was resulting in an ongoing cycle of churnout due to workload instability.
Startup Weekend 2017 - 1st Place Winner
At Startup Weekend 2017 our team designed a solution for extending internet connectivity out at
sea through a network
of buoys acting as cell towers, allowing oil rigs to stay on top of the latest updates for
iceberg locations and take
preventitive action.
Newfoundland Developers (NDev) Talk - September 2022
At this talk I presented some of the theory behind Microserviced architectures, and common
misconceptions
in practice within the industry on how and when they should be implemented.
Experience
Mysa
Software Architect (Cloud) - St. John's, NL, Canada
Aug. 2019 - Present (4+ years)
- Designed a lightweight and novel approach for replicating, ETLing, cataloging and performing extensive and complex queries on Terabytes of DynamoDB data at almost no cost
- Automated and unified Mysa's manual cloud deployments, by developing a custom framework and tooling for orchestrating deployments across different microserviced cloud applications in tandem, using Lerna, PNPM, GitHub API and AWS CDK
- Invented an industry leading droop correction algorithm for Mysa's thermostats, that makes it one of the very best line-voltage thermostats in terms of high-load thermal accuracy
- Developed an Express-inspired framework over AWS CDK that significantly streamlines the complexity of developing serverless API endpoints on multiple units of infrastructure
- Devised an approach to achieve 100% monitoring coverage of 1000s of lambda functions within days, at near-zero cost
Google
Software Engineering Intern - Kirkland, WA, US
Apr. 2019 - Jul. 2019 (3 Months)
- Constructed an automated canonical definition of multiple sources of code, by implementing a language parser based on an Abstract Syntax Tree in Dart, and integrating it with a filesystem watcher
- Decreased the running time of integration tests from ~2h30mins to ~30mins, by capitalizing on the caching of repeated operations across parallel tests
Mysa
Software Developer - St. John's, NL, Canada
Jul. 2018 - Apr. 2019 (10 Months)
- Pioneered a machine learning pipeline using Scikit-learn, and ran it on AWS Batch, Step Functions and DynamoDB, to quarter the relative error of time-to-setpoint prediction from 58% to 13% and boost energy savings
- Developed droop correction models, by programming Arduino loggers for data collection, aggregating data using Pandas, and applying feature+model selection using Scikit-learn
Memorial University of Newfoundland
Undergraduate Research Assistant - St. John's, NL, Canada
May 2018 - Aug 2018 (4 Months)
- Investigated enhancements to generative ML models that increase entropy of output distributions, by applying Boosting techniques to GANs
- Proposed novel methods of encoding sequential data, such as music, in a form that optimises for GANs and other non-sequential models
HeyOrca!
Junior Web Developer - St. John's, NL, Canada
May 2017 - Jun 2018 (1 year)
- Quadrupled the speed of critical services through managed asynchronization of Laravel queues, optimization of data structures and logic refactoring
- Rebuilt and enhanced the frontend using ReactJS, along with Web Sockets and REST APIs, in order to create a more performant and resilient SPA
- Reduced the SPA’s bundle size by ~40%, through proper management of dependencies and Webpack builds for all frontend micro-services
Certifications & Publications
AWS Solutions Architect - Associate
Awarded May 2021
Earners of this certification have a comprehensive understanding of AWS services and technologies. They demonstrated the ability to build secure and robust solutions using architectural design principles based on customer requirements. Badge owners are able to strategically design well-architected distributed systems that are scalable, resilient, efficient, and fault-tolerant.
Improving Bacterial sRNA Identification By Combining Genomic Context and Sequence-Derived Features
Computational Intelligence Methods for Bioinformatics and Biostatistics (CIBB) - 2021
Mohammad Sorkhian, Megha Nagari, Moustafa Elsisy & Lourdes Peña-Castillo
Bacterial small non-coding RNAs (sRNAs) are ubiquitous regulatory RNAs involved in controlling several cellular processes by targeting multiple mRNAs. The large diversity of sRNAs in terms of their length, sequence, and function poses a challenge for computational sRNA prediction. In this paper, we combine ensemble learning over genomic context features with sequence-derived features to invent a model with demonstrably higher predictive performance according to measurements of AUPRC against current state of the art models.
Awards
MUN University Medal of Academic Excellence - Computer Science
This medal is awarded to a student of the highest level of academic achievement in a convocating class of the Department of Computer Science. This medal has been awarded with a cumulative academic average of 96.5%
MUN Faculty of Science Book Prize - Computer Science 2017
Each year, every department in the Faculty of Science nominates a student who is not graduating in that year to win this award based on demaonstrated academic excellence.
MUN Faculty of Science Dean's List
Hackathons Won
Projects
GANs 
Reels
Checkout this musical piece by a Convolutional Neural Network! At the GANs n Reels project, we encoded sequential data, such as music, into spatial data in order to investigate whether such encoding would help a CNN learn relationships within and across bars of a musical piece! This would then be used in a Generative Adverserial Network to generate novel pieces of music :) If you would like to learn more about GANs n Reels, checkout our project page
here where you can listen to the actual pieces, or checkout our ArXiv paper here.
ML Rock-Paper-Scissors
This is a lightweight project that compares a number of common statistical, procedural and Deep Learning algorithms. It started off as a web-based
demo, but I last left off in the middle of a migration to Python, since that was giving me better access to statistical and ML libraries. I am looking
at regenerating and posting the results from this experient, but in the meantime you can find the source code for this project here.
Education
Memorial University of Newfoundland
B.Sc. (Hons), Computer Science
2016-2019
GPA: 4.0
Cumulative Average: 96.5%
HackFrost 3.0 Hackathon - 1st Place Winner
HackFrost 3.0 was a hackathon targetted at tackling emissions from industrial transportation. Our team, Viridis, designed a DIY kit for retrofitting long-haul trucks with solar-powered Auxilliary Power Units (APUs) - a solution that is near-immediately deployable, scalable and with the potential to annually offset 1.5x NYC's annual emissions every year if deployed across North America. Over 150 people participated at the event.
Mysa's Group Photo - 2019
Mysa's group photo in 2019 (with me ecstatically pointing at the Mysa logo 😂)
Googleplex - Google CodeU Retreat (2018)
Google CodeU is a 12 week invitation-only virtual program, composed of 150 high potential undergraduates pursuing Computer Science in North America. At the end of the program we were hosted over for a 3-day retreat at Google's Headquarters in Mountain View :)
Bounce Hackathon - 2023
The 2023 Bounce Hackathon was targetted at addressing employment retention issues in the health sector within NL. As part of this Hackathon, I developed a rough simulation of the impact of employee churnout on the workload of remaining team members, and how the current rate of churnout relative to employment was resulting in an ongoing cycle of churnout due to workload instability.
Startup Weekend 2017 - 1st Place Winner
At Startup Weekend 2017 our team designed a solution for extending internet connectivity out at sea through a network of buoys acting as cell towers, allowing oil rigs to stay on top of the latest updates for iceberg locations and take preventitive action.
Newfoundland Developers (NDev) Talk - September 2022
At this talk I presented some of the theory behind Microserviced architectures, and common misconceptions in practice within the industry on how and when they should be implemented.
Experience
Mysa
Software Architect (Cloud) - St. John's, NL, Canada
Aug. 2019 - Present (4+ years)
- Designed a lightweight and novel approach for replicating, ETLing, cataloging and performing extensive and complex queries on Terabytes of DynamoDB data at almost no cost
- Automated and unified Mysa's manual cloud deployments, by developing a custom framework and tooling for orchestrating deployments across different microserviced cloud applications in tandem, using Lerna, PNPM, GitHub API and AWS CDK
- Invented an industry leading droop correction algorithm for Mysa's thermostats, that makes it one of the very best line-voltage thermostats in terms of high-load thermal accuracy
- Developed an Express-inspired framework over AWS CDK that significantly streamlines the complexity of developing serverless API endpoints on multiple units of infrastructure
- Devised an approach to achieve 100% monitoring coverage of 1000s of lambda functions within days, at near-zero cost
Software Engineering Intern - Kirkland, WA, US
Apr. 2019 - Jul. 2019 (3 Months)
- Constructed an automated canonical definition of multiple sources of code, by implementing a language parser based on an Abstract Syntax Tree in Dart, and integrating it with a filesystem watcher
- Decreased the running time of integration tests from ~2h30mins to ~30mins, by capitalizing on the caching of repeated operations across parallel tests
Mysa
Software Developer - St. John's, NL, Canada
Jul. 2018 - Apr. 2019 (10 Months)
- Pioneered a machine learning pipeline using Scikit-learn, and ran it on AWS Batch, Step Functions and DynamoDB, to quarter the relative error of time-to-setpoint prediction from 58% to 13% and boost energy savings
- Developed droop correction models, by programming Arduino loggers for data collection, aggregating data using Pandas, and applying feature+model selection using Scikit-learn
Memorial University of Newfoundland
Undergraduate Research Assistant - St. John's, NL, Canada
May 2018 - Aug 2018 (4 Months)
- Investigated enhancements to generative ML models that increase entropy of output distributions, by applying Boosting techniques to GANs
- Proposed novel methods of encoding sequential data, such as music, in a form that optimises for GANs and other non-sequential models
HeyOrca!
Junior Web Developer - St. John's, NL, Canada
May 2017 - Jun 2018 (1 year)
- Quadrupled the speed of critical services through managed asynchronization of Laravel queues, optimization of data structures and logic refactoring
- Rebuilt and enhanced the frontend using ReactJS, along with Web Sockets and REST APIs, in order to create a more performant and resilient SPA
- Reduced the SPA’s bundle size by ~40%, through proper management of dependencies and Webpack builds for all frontend micro-services
Certifications & Publications
AWS Solutions Architect - Associate
Awarded May 2021
Earners of this certification have a comprehensive understanding of AWS services and technologies. They demonstrated the ability to build secure and robust solutions using architectural design principles based on customer requirements. Badge owners are able to strategically design well-architected distributed systems that are scalable, resilient, efficient, and fault-tolerant.
Improving Bacterial sRNA Identification By Combining Genomic Context and Sequence-Derived Features
Computational Intelligence Methods for Bioinformatics and Biostatistics (CIBB) - 2021
Mohammad Sorkhian, Megha Nagari, Moustafa Elsisy & Lourdes Peña-Castillo
Bacterial small non-coding RNAs (sRNAs) are ubiquitous regulatory RNAs involved in controlling several cellular processes by targeting multiple mRNAs. The large diversity of sRNAs in terms of their length, sequence, and function poses a challenge for computational sRNA prediction. In this paper, we combine ensemble learning over genomic context features with sequence-derived features to invent a model with demonstrably higher predictive performance according to measurements of AUPRC against current state of the art models.Awards
MUN University Medal of Academic Excellence - Computer Science
This medal is awarded to a student of the highest level of academic achievement in a convocating class of the Department of Computer Science. This medal has been awarded with a cumulative academic average of 96.5%
MUN Faculty of Science Book Prize - Computer Science 2017
Each year, every department in the Faculty of Science nominates a student who is not graduating in that year to win this award based on demaonstrated academic excellence.
MUN Faculty of Science Dean's List
Hackathons Won
Projects
GANs Reels
Checkout this musical piece by a Convolutional Neural Network! At the GANs n Reels project, we encoded sequential data, such as music, into spatial data in order to investigate whether such encoding would help a CNN learn relationships within and across bars of a musical piece! This would then be used in a Generative Adverserial Network to generate novel pieces of music :) If you would like to learn more about GANs n Reels, checkout our project page here where you can listen to the actual pieces, or checkout our ArXiv paper here.
ML Rock-Paper-Scissors
This is a lightweight project that compares a number of common statistical, procedural and Deep Learning algorithms. It started off as a web-based demo, but I last left off in the middle of a migration to Python, since that was giving me better access to statistical and ML libraries. I am looking at regenerating and posting the results from this experient, but in the meantime you can find the source code for this project here.
Education