Ferhat Abdullahoglu

I was involved in this project from the design phases until it had over ten updates on the App Store. That means I took part in deciding the architecture the app would have based on the requirements. We used Swift Package Manager to help us have a modular, scalable, and testable codebase. More than 100 modules were brought together to provide the required functionality. Composable architecture (TCA) was used for the architecture, and the UI was built using SwiftUI.

Joined this project while it was still in an MVP phase. I helped finish the remaining features faster. Then throughout multiple releases, I've continued adding more features while at the same time optimizing and debugging the existing codebase to improve the app code quality and the user experience.

I worked on the full-stack development of a chatbot app that generates personalized mindfulness meditation audios based on the conversation with the user. I developed the front and back end from scratch for the App Store release.
The server app used GPT-4o and Llama 3.3 to handle conversations with the user and generate scripts. The iOS app was developed with SwiftUI. The app is one of a kind, and I delivered the project successfully.

This Swift-based app delivers mindfulness meditations from different categories. It comes with a companion watch app through which users can select their favorite meditations and stream them directly. Additionally, users can choose affirmations from different categories to focus on and create a new meditation recording on the spot out of these selected affirmations.

This was made possible by AWS and our custom back-end systems working together. The back-end processes the meditation request, which sends a request to AWS Elemental MediaConvert, using the AWS API Gateway to create a new meditation session. Finally, the app gets this information from the back end to serve the content to the user. We also implemented the Web2App concept, a feature that administers a quiz on https://premium.crea.care and allows the creation of a specific set of meditations based on the answers.

Built a RAG solution that is still under active development and accessible here: streamlitalb-1496149066.eu-central-1.elb.amazonaws.com. The entire solution flow is as follows:

• Crawling a company's public webpage
• Using LLama over Amazon Bedrock to identify which links are related to ESG
• Scraping text content, downloading PDF docs, and storing everything on AWS with a combination of DynamoDB and Amazon S3
• Generating embeddings using text-embedding-ada-002 from OpenAI
• Storing resulting vectors at Pinecone
• Switching the context based on the selected company and answering questions through RAG using GPT-4

An open source API client wrapper around Alamofire written with Swift, leveraging Swift concurrency APIs.

It's especially helpful as it still uses Alamofire, which is a very well-known framework, in its backbone and enables defining fully typed routes and executing the client from within an async scope

This native iOS app was built to bring a good quality pilates workout experience to people's homes. It uses Siri shortcuts and voice commands, AirPlay, a companion watch app, and Chromecast to create an uninterrupted workout experience. The app integrates with a CDN and a back-end application to deliver smooth streaming (HLS) performance and to sync users' data among all their devices. The app also integrates with HealthKit to update the log user's statistics if the user chooses to do so.

A Swift framework for implementing a story functionality similar to Instagram stories. It was developed as part of a project, then made public.
It makes use of Generics to support different content types.

A custom GPT backed by a back-end component that creates mindfulness meditation audio records within ChatGPT.
The back end makes use of these components from AWS:
• SAM
• CloudFront
• CloudFormation
• Route53
• Lambda
• API Gateway
• Secrets Manager
• S3