We can organize an interview with Aldin or any of our 25,000 available candidates within 48 hours. How would you like to proceed?
Schedule Interview NowAt Softaims, I have been fortunate to work in an environment that values creativity, precision, and long-term thinking. Each project presents a unique opportunity to transform abstract ideas into meaningful digital experiences that create real impact. I approach every challenge with curiosity and commitment, ensuring that every solution I design aligns not just with technical requirements, but also with human needs and business objectives. One of the most rewarding aspects of my journey here has been learning how to bridge the gap between innovation and practicality. I believe technology should simplify complexity, enhance efficiency, and empower people to do more with less friction. Whether building internal systems, optimizing workflows, or helping bring client visions to life, my focus remains on developing solutions that stand the test of time. Softaims has encouraged me to grow beyond coding—to think about design, communication, and sustainability in technology. I see every project as part of a larger ecosystem, where small details contribute to long-lasting results. My daily motivation comes from collaborating with people who share the same passion for doing meaningful work, and from seeing the tangible difference our efforts make for clients around the world. More than anything, I value the culture of learning and improvement that defines Softaims. It’s a place where ideas evolve through teamwork and constructive feedback. My goal is to continue refining my craft, exploring new approaches, and contributing to solutions that are not only efficient but also elegant in their simplicity.
Main technologies
3 years
2 Years
1 Year
2 Years
Potentially possible
This project's primary objective was to generate illustrations for children's books. For this purpose, I utilized and tested some different checkpoints with combinations of various LoRAs to get results as clients need.
During this project, our main goal was to revolutionize the appearance of QR codes by infusing them with captivating styles. To achieve this, we leveraged ControlNet's powerful InPaint Preprocessor in conjunction with two distinct ControlNet Models: the Brightness Model and the Tile Model. These models worked harmoniously to generate stunning images of QR codes that maintained their scannability while exuding artistic flair. By integrating advanced AI techniques seamlessly, we transformed QR codes into visually appealing elements that captured users' attention. This project showcased our expertise in ControlNet technology and highlighted our commitment to pushing the boundaries of innovation in image generation and customization. The knowledge regarding the parameters of Stable Diffusion and ControlNet came in handy to produce a refined image. After careful adjustments and trial and error, this project allowed us to get the best settings for Stylistic QR generation themes.
During my tenure at DevX House in Faisalabad, Pakistan, I had the opportunity to work as a Flutter App Developer on an exciting project that aimed to create a music streaming app similar to Spotify. Leveraging my Flutter, Dart, and FireBase expertise, I played a key role in developing a robust and feature-rich application. One of my work's primary focuses was crafting a highly responsive user interface that closely resembled the intuitive design of Spotify. By meticulously designing and implementing the UI components, I ensured an engaging and immersive experience for users, allowing them to navigate and discover music effortlessly. To enhance the overall user experience, I integrated the Provider Pattern, a state management solution, which facilitated efficient data flow and seamless updates across the app. This pattern optimized performance and enabled smooth transitions between different screens and interactive features. To keep users informed and engaged, I implemented FireBase Functions to send relevant push notifications efficiently. By leveraging the power of FireBase Cloud Messaging, I ensured that users received timely updates about new releases, personalized playlists, and recommended tracks, contributing to an enhanced user engagement rate. For managing the vast amount of music data, I utilized FireStore DataBase, creating collections to store user profiles, song metadata, playlists, and artist information. By establishing relationships and leveraging the querying capabilities of FireStore, I enabled seamless browsing and searching of songs, albums, and playlists, providing a comprehensive music discovery experience. In addition to data management, I provisioned FireBase Cloud Storage to store and retrieve mp3 files securely. This allowed users to stream their favorite songs seamlessly, ensuring a smooth and uninterrupted music playback experience. Throughout the project, I collaborated closely with a cross-functional team, including designers and backend developers, to ensure a cohesive and high-quality final product. Together, we successfully developed a music streaming app that mirrored the functionality and aesthetics of Spotify, offering users a seamless platform to explore and enjoy their favorite music. Working on this Spotify-inspired music app was an incredibly rewarding experience, as it allowed me to showcase my expertise in Flutter app development, Dart programming, and FireBase integration. I take pride in my contribution to delivering a top-notch music streaming app that brings the joy of music to users' fingertips.
This project's primary objective was to develop a robust system capable of generating AI avatars and images based on user input and chosen styles. The project revolved around a user-centric approach, where the user provided about 10 to 16 images as references and selected a specific style for the generated output. Using Stable Diffusion, I designed and developed models to accomplish this task efficiently. Here are the key highlights of the project: Data Preparation: The input data was prepared by adjusting its resolution to 512 X 512, which is ideal for training such models. Denoising was also done on some images to improve them, and the images that could have a negative effect on the model training were removed, as even a single picture can heavily influence the Model. Kohya Dreambooth LoRA Model: This is the first Model generated after carefully adjusting the parameters and understanding what is happening in the backend. There is no shame in admitting it took some trial and error to produce a decent model. Dreambooth Checkpoint Model Generation: Although this wasn't part of the original plan, I wanted to improve the Model's output to give my client an excellent Model; I moved towards training a full-fledged Stable Diffusion Checkpoint Model. It took a while; we made relevant adjustments to the parameters, and the results seemed much better. LoRA Extracted from Checkpoint: The reason for this step was to not only reduce the size of the Model but also be able to use it in conjunction with other checkpoints which are trained for producing realistic images. This Model also performed very well. Improved LoRA Model: This last step successfully attempted to improve the Kohya Dreambooth LoRA Model. We slightly changed the parameters and started to receive promising results. The client was looking to make a method where users will upload a few photographs of themselves, and a Model will be trained, which will output their images in different styles. I needed to provide my client with an efficient solution to achieve this, so I went the extra mile to get their desired results filled by directly training a LoRA model. This improves efficiency in two significant ways, it's quicker to train and takes lesser space. X/Y/Z Plotting: After training, evaluating the resulting models is vital to find the best combination of steps and LoRA weight. The best way to assess the Models is using the x/y/z plot. I used this technique to assess which Model is working best. Regularization Images: We used AI-Generated and Professional Photographs of people to train the Model to assess what will work the best. (Spoiler Alert: it's the AI-Generated Images). Prompts: We generated a couple of negative prompts, which will work fine in any scenario related to the client's needs. We also made a couple of other prompts and guided the client with the best ways to do so. I prioritized delivering exceptional quality and fine-grained control over the generated output throughout the project. I successfully fulfilled the client's requirements by combining my expertise in AI/ML with diligent model development and effective style selection.
Designed this newsletter cover on urgent needs base in under 1.5 hours, turned out pretty good.
Bachelor of Science (BS) in Computer science
2020-01-01-2024-01-01