(MAY 2023 - Present)
Machine Learning Engineer
Cilans System
AI_ML_engineer={name:'Parth Panchal',skills:['Python', 'Machine Learning', 'Deep Learning', 'Natural Language Processing', 'TensorFlow', 'PyTorch', 'Scikit-learn', 'Pandas', 'NumPy','Azure AI'],passionate:True,analytical_thinker:True,problem_solver:True,hireable:lambda(AI_ML_engineer['passionate']&AI_ML_engineer['problem_solver']&len(AI_ML_engineer['skills'])>=5)}ABOUT ME
Who I am?
I enjoy working in a role that offers diverse challenges, fosters innovation, and allows me to collaborate on cutting-edge AI/ML projects. I thrive in environments where I can continuously learn and contribute to impactful solutions.
Experiences
(NOV 2022 - MAY 2023)
Data Science Intern
Skills
PROJECTS
Lookr.fyi (US-based Startup)
project={name:'Lookr.fyi (US-based Startup)',tools: ['Azure AI', 'Azure Machine Learning', 'Azure Openai', 'Hugging-face', 'RAG', 'Prompt engineering', 'Python],myRole:AI/Ml Engineer,Description: Developed an AI-driven search engine for fashion products, enabling user-centric query-based discovery through a RAG-based deep search solution integrating text, vector, and image search capabilities. Leveraged Azure AI Search, Azure Machine Learning, and PromptFlow to design an efficient and scalable system. The solution incorporated Azure Functions for microservices, Cosmos DB for vector storage, and Blob Storage for seamless log management. Deployed and managed the complete life cycle on Azure, ensuring robust production operations while driving ongoing enhancements to optimize system performance.,};Jewel.AI
project={name:'Jewel.AI',tools: ['Tensorflow', 'OpenCV', 'Replicate Ai', 'Openai Api', 'Hugging-face', 'Stable-diffusion', 'Prompt engineering', 'LoRA', 'Confluence', 'Git', 'Python],myRole:AI/Ml Engineer,Description: Me and my team build an AI application tailored for the jewelry industry, automating image-based description generation. This involved fine-tuning a unique jewelry image generator model using stable diffusion with LoRA. To enhance functionality, I integrated various open-source models into the system. I also fostered a collaborative environment across front-end, back-end, and DevOps teams, streamlining integration processes for efficient deployment.,};Failure Prediction of Compressor in Oil-Rig
project={name:'Failure Prediction of Compressor in Oil-Rig ',tools: ['Tensorflow', 'Keras', 'LSTM', 'Python', 'Pandas', 'Numpy', 'Scikit-learn', 'Matplotlib],myRole:Machine Learning Engineer,Description: Spearheaded the development of a predictive maintenance project focused on compressor systems, achieving a significant enhancement in accuracy from 84% to 90% with a 3-4 member team. Using TensorFlow and LSTM, I led the implementation of advanced predictive modeling techniques. Additionally, I applied rigorous feature engineering, data cleaning, and preprocessing methods including PCA, ridge, and lasso to optimize model performance.,};Automated Document Information Extraction System
project={name:'Automated Document Information Extraction System',tools: ['Azure AI', 'Azure Machine Learning', 'Azure Openai', 'Colpali ', 'PyTesseract],myRole:AI/Ml Engineer,Description: Architected and deployed an end-to-end pipeline to extract structured data from financial documents into pre-defined Excel templates, leveraging Azure OpenAI models for automated data extraction with scalable cloud integration. Led a team of 3–4 developers, managing project workflows and collaborating closely with clients to refine requirements and deliver impactful solutions. Implemented ColBERT-based retrieval models for multi-page PDF parsing and deployed the solution on Azure Machine Learning for robust performance.,};Face Recognition Project with MLOps
project={name:'Face Recognition Project with MLOps',tools: ['TensorFlow', 'MlFlow', 'DVC', 'Git-hub Actions', 'Docker', 'OpenCV', 'AWS', 'Dagshub', 'Tensorboard', 'EC2', 'ECR],myRole:Machine Learning Engineer,Description: This is the project where I showcased my expertise in model building, focusing on TensorFlow and VGG16 architecture for transfer learning to develop advanced deep learning models. I integrated deep learning techniques using openCV and haarcascade for precise face detection applications. To ensure efficient development and deployment, I incorporated MLOps best practices, utilizing MLflow for model tracking, DVC for data versioning, and Dagshub for collaboration and experiment management. For scalable cloud-based deployment on AWS, I implemented Docker, leveraging TensorBoard for visualization and customized log structures to enhance debugging efficiency.,};Deep Learning-driven Text Summarization
project={name:'Deep Learning-driven Text Summarization',tools: ['PyTorch', 'Docker', 'AWS', 'FASTAPI', 'IAM', 'EC2', 'ECR', 'GitHub Actions],myRole:Machine Learning Engineer,Description: In this project I developed a text summarization solution using PyTorch and Google’s Pegasus model, leveraging deep learning principles in NLP with advanced architectures like Transformers. I utilized Docker for containerized deployment, demonstrating scalability and operational efficiency. Furthermore, I implemented AWS CI/CD deployment through GitHub Actions, encompassing IAM user creation, ECR repository setup, EC2 machine configuration, and Docker installation. This project underscores my skills in AI, NLP, containerization, and cloud deployment, reflecting my ability to deliver robust and scalable solutions in machine learning-driven applications.,};SocialSyncAgent – An Agentic AI System
project={name:'SocialSyncAgent – An Agentic AI System',tools: ['Python', 'Langchain', 'Langgraph', 'LAngsmith', 'Gemini', 'Agentbox],myRole:AI/ML Engineer,Description: An AI-powered automation system designed to revolutionize content curation and social media management. This intelligent agent aggregates data from multiple sources—blogs, tweets, GitHub repositories, and more—then analyzes and synthesizes the information into comprehensive content reports. It automatically adapts and formats posts to suit each platform, ensuring character limits and stylistic guidelines are met for LinkedIn, Twitter/X, and others.Featuring smart scheduling capabilities, it allows users to set customized posting dates and priorities while providing a human-in-the-loop feedback system for content review and approval. With direct integration for automated multi-platform publishing, the Social Media Agent simplifies content distribution, making it efficient and seamless.,};Certifications
