π‘ Hello there! I am Parth
I am a passionate AI Software Developer with expertise in machine learning, deep learning, computer vision, and generative AI. My work focuses on creating innovative, data-driven solutions that optimize industrial processes and deliver real-world impact. From developing vision transformer architectures to deploying scalable AI models, I thrive at the intersection of cutting-edge technology and practical applications. Currently, at ISRA Vision GmbH (part of Atlas Copco), I build advanced AI solutions for industrial automation, leveraging tools like PyTorch, TensorFlow, and Kubernetes to drive efficiency and scalability.
π Featured Projects
πΌοΈ Vision Transformer for Synthetic Data Generation
Leading a generative AI initiative at ISRA Vision:
- π― Vision: Developing vision transformer architectures for synthetic data generation
- π‘ Innovation:
- Implemented advanced deep learning models for data synthesis
- Integrated rule-based decision agents for automated workflows
- Reduced data acquisition costs by 30% through generative AI
- π Impact:
- Enabled scalable AI training with synthetic datasets
- Streamlined industrial inspection processes
- Enhanced model robustness for real-world applications
π Defect Detection for Car Surface Inspection
Spearheading an AI-driven quality control project at Atlas Copco:
- π― Vision: Enhancing car surface inspection through deep learning
- π‘ Innovation:
- Developed U-Net and ResNet-based models for defect detection
- Incorporated adaptive gamma correction for image enhancement
- Achieved 70% structural similarity index (SSIM) and 10% accuracy improvement
- π Impact:
- Reduced labor costs by 90% through automated inspection
- Processed 800GB+ image datasets for precise 2D/3D mapping
- Improved quality control efficiency in automotive manufacturing
π Recent Blog Posts
πΌοΈ Building a CLIP-like Model for Image-Text Alignment
Sharing insights on multimodal AI development:
- π― Vision: Guiding developers to build a CLIP-like model for image-text alignment
- π‘ Innovation:
- Step-by-step tutorial using PyTorch and Hugging Face transformers
- Covers data preparation, model training, and evaluation for multimodal learning
- Demonstrates practical implementation of vision-language models
- π Impact:
- Empowers developers to create AI systems bridging images and text
- Enhances understanding of transformer-based architectures
- Published on Medium, April 2025
π― Research Interests
AI and Machine Learning:
- π€ Deep Learning and Computer Vision
- π Generative AI and Vision Transformers
- π MLOps and Scalable AI Deployment
- π Time Series Forecasting with Transformers
Industrial Applications:
- π Industrial Automation and Quality Control
- π Defect Detection and Image Processing
- π‘ Large-Scale Data Processing
- βοΈ Sustainable AI Optimization
π Recent Achievements
- IEEE Publication (2023)
- π Published research on data-driven deep learning for Bitcoin price forecasting
- π Leveraged transformer-based models for 22% accuracy improvement over ARIMA
- Deutschlandstipendium Scholarship
- π Awarded for academic excellence at Technical University of Darmstadt
- π Top performer in Information and Communication Engineering
- TCS CodeVita Qualifier
- π» Recognized for competitive programming and algorithmic problem-solving
- π§ Demonstrated proficiency in data structures and algorithms
π Education
- M.Sc. in Information and Communication Engineering
- ποΈ Technical University of Darmstadt (GPA: 2.0/5.0, ~3.5/4.0 equivalent)
- π Deutschlandstipendium Holder
- π¬ Relevant Coursework: Machine Learning, Deep Learning, Image Processing, Data Structures and Algorithms
- B.Eng. in Electronics and Communication Engineering
- π Vishwakarma Government Engineering College, Ahmedabad (GPA: 8.8/10)
- π Focus: Signal Processing and Embedded Systems
πΌ Professional Experience
AI Software Developer, ISRA Vision GmbH (Atlas Copco), Germany (Oct 2023 β Jun 2025)
- π Develop vision transformer models for generative AI and industrial automation
- πΎ Deploy scalable AI solutions using Docker and Azure Kubernetes Service
- β‘ Optimize workflows with rule-based agentic systems, reducing costs by 30%
Software Development Engineer, Atlas Copco, Germany (Mar 2022 β Sep 2023)
- πΌοΈ Built U-Net and ResNet models for car surface defect detection, achieving 90% labor cost reduction
- π Processed 800GB+ datasets for 2D/3D mapping and quality control
- π§ Enhanced model accuracy by 10% with adaptive image processing techniques
Data Science Intern, Spryfox GmbH, Darmstadt (Oct 2021 β Mar 2022)
- π Developed machine learning models (XGBoost, Logistic Regression) for pet disease prediction, improving accuracy by 20%
- π» Engineered end-to-end data science pipeline with AWS deployment
- π Conducted statistical analysis for customer spending insights
π― Teaching and Outreach
- YouTube Content Creator
- π₯ Create educational content on machine learning, data science, and programming
- π₯ Community of 3.9k+ subscribers
- π Tutorials on Python, PyTorch, and deep learning applications
- DAAD Student Representative
- π Technical University of Darmstadt
- π€ Led cross-cultural communications for 100+ international students
- π Organized tech-focused workshops and events
π Current Learning Journey
Udemy Courses In Progress
| Course Name | Instructor | Progress | Certificate |
|---|---|---|---|
| End-to-End MLOps Bootcamp: Build, Deploy, and Automate ML with Data Science Projects | Krish Naik | 40% | π In Progress |
| Complete Guide to Building, Deploying, and Optimizing Generative AI with Langchain and Huggingface | Krish Naik | 30% | π In Progress |
