I’m a software engineer with experience in software development for machine learning research and web development.
Leveraging Synthetic Images with Domain-Adversarial Neural Networks for Fine-Grained Car Model ClassificationI implemented and evaluated domain generalization algorithms for leveraging synthetic images for fine-grained car model classification.
I used DomainBed for performing the experiments.
Ended my journey as Machine Learning Engineer at SAP!
New bahá’í song project websiteThe bahá’í song project website provides videos, lyrics and chords for Bahá’í-inspired songs.
I have maintained bahá’í song project since 2011 growing the YouTube channel to 3000+ subscribers, and the website to 1300+ monthly visitors and 8000+ monthly page views.
The last big change was the development of a new website. The website was originally based on WordPress, which came with a few challenges: continuous development was difficult, maintaining 100+ songs in WordPress posts became infeasible since data was often duplicated and there was not one single source of truth. I migrated the existing data to a relational database and designed and implemented a GraphQL API with Prisma. Song sheets are maintained in text files that serve as the source files for PDFs generated with ChordPro. The GraphQL API together with GatsbyJS, Tailwind CSS and Netlify enabled me to quickly develop and deploy a new, high-performing website that is easy to maintain.
bahá’í song project is open source and the repositories are available on GitHub.
I collaborated with another software engineer to try to understand if the size of wildfires can be forecast with deep learning using meteorological data. The experiments showed that meteorological data alone does not seem to provide enough information, but is valuable for augmenting other data such as satellite images.
Market research for a smart lighting system startup
The team was asked to validate the smart lighting system for a specific market segment. Through a survey and interviews with executives in this segment, the team could recommend a specific subsegment for which the lighting system was most fitting.
I did quantitative market research, designed the survey and the interview guide, gave weekly status updates and presented the recommendations to the client.
Near-real-time lightning analysis
I collaborated with another software engineer to visualize and analyze near-real-time lightning data from the Lightning Imaging Sensor on the International Space Station.
I set up the ETL pipeline to provide easy access to the NASA Earthdata lightning data.