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 NowMy name is Shehryar M. and I have over 4 years of experience in the tech industry. I specialize in the following technologies: Python, Computer Vision, TensorFlow, OpenCV, PyTorch, etc.. I hold a degree in High school degree, Master of Computer Science (MSCS), Bachelor of Science (BS). Some of the notable projects I’ve worked on include: LLM-Based Conversational AI System, Fine-tune Stable Diffusion using LoRA for Doodle Generation, AI-Generated Headshots using Flux and LoRA, Soccer Analytics using Person Detection and Tracking Models, License Plate Detection using YOLO and DeepSort, etc.. I am based in Lahore, Pakistan. I've successfully completed 8 projects while developing at Softaims.
I approach every technical challenge with a mindset geared toward engineering excellence and robust solution architecture. I thrive on translating complex business requirements into elegant, efficient, and maintainable outputs. My expertise lies in diagnosing and optimizing system performance, ensuring that the deliverables are fast, reliable, and future-proof.
The core of my work involves adopting best practices and a disciplined methodology, focusing on meticulous planning and thorough verification. I believe that sustainable solution development requires discipline and a deep commitment to quality from inception to deployment. At Softaims, I leverage these skills daily to build resilient systems that stand the test of time.
I am dedicated to making a tangible difference in client success. I prioritize clear communication and transparency throughout the development lifecycle to ensure every deliverable exceeds expectations.
Main technologies
4 years
1 Year
1 Year
1 Year
Potentially possible
Techlogix
The goal was to develop a conversational AI capable of reading and analyzing complex Offering Memorandum documents and answering user queries: - Used LLM with Langchain to build a Retrieval-Augmented Generation (RAG) system. - Integrated OCR to extract text from documents, including tables and images. - Employed Document Layout Parser to maintain document structure and layout. - Incorporated text-to-speech (TTS) and speech-to-text (STT) for seamless voice-based interactions. - The AI could parse and understand complex documents, providing accurate answers and insights based on user queries.
The goal was to fine-tune Stable Diffusion to generate sketch doodle-style images of characters in various poses and scenes: - Scraped hundreds of doodle images from the web using Selenium. - Generated captions for images with BLIP to aid text-to-image training. - Trained a baseline Stable Diffusion model locally with these images and captions. - Fine-tuned the model using LoRA to produce consistent character images in different poses. - Deployed the final model on Runpod.io and exposed it via an API.
Develop AI-generated headshots using Flux, with fine-tuning and prompt engineering for varied poses: - Fine-tuned Flux on a person’s image using LoRA (Low-Rank Adaptation), which efficiently adapts models to specific tasks. - Applied prompt engineering to generate headshots in multiple poses, expressions and angles. - Included automated quality checks to ensure user-uploaded photos are of high quality - Created a FastAPI application to facilitate easy access and integration. - Deployed the model on AWS servers and built an automated pipeline to train on user images and generate results.
A client needed an AI tool to analyze soccer matches and provide insights for player improvement: - Trained YOLO-based model to detect players and the ball using annotated match videos. - Used DeepSort to track players and the ball throughout the match. - Developed a clustering technique to re-identify players when they re-entered the camera view. - Built a model to detect team affiliation based on shirt colors. - Implemented event detection to identify passes, intercepts, and dribbles for in-depth player analysis.
The goal was to build a license plate recognition system for real-time video analysis: - Used YOLO for detecting license plates, Deep SORT for tracking, and OCR for reading plate numbers. - Trained YOLO on a custom dataset to improve accuracy across different angles and lighting conditions. - Achieved real-time performance with video processing at 10 frames per second (fps). - Optimized the system for efficient and accurate detection and tracking. - Ideal for live surveillance and automated vehicle monitoring.
High school degree in
2004-01-01-2015-01-01
Master of Computer Science (MSCS) in Computer science
2019-01-01-2021-01-01
Bachelor of Science (BS) in Electrical engineering
2015-01-01-2019-01-01