Mit Patel

Assistant Lecturer at Dorset College

2022 - Present

Taught 300+ undergraduate students across Python, Data Analysis, Machine Learning, and Statistics, achieving 92-96% project completion rates and 88-90% course pass rates through practical, student-focused pedagogy.

• Designed and delivered industry-aligned curricula, regularly updating modules to reflect modern trends while ensuring accessibility for diverse academic backgrounds.
• Led comprehensive Python and web development sessions using Flask, guiding students through building real applications (blog platform, banking system) covering environment setup, routing, templating, database integration with SQLAlchemy, and authentication.
• Created structured project-based learning flow from theoretical foundation to hands-on implementation, integrating Git/GitHub for version control and culminating in end-to-end capstone projects that simulate real-world development workflows.
• Conducted ML training sessions emphasizing classical algorithms, evaluation metrics, and real-world applications using Scikit-learn, Pandas, NumPy, and PyTorch, supporting students through guided statistical modeling projects.
• Initiated cloud technology exposure through AWS Educate classes and integrated external platforms like DataCamp to enhance independent skill development beyond classroom instruction.
• Mentored students on industry best practices including modular coding, documentation, reproducibility, and professional communication while promoting code clarity and testing without overwhelming beginners.
• Integrated external learning platforms like DataCamp into the curriculum to support independent skill development and hands-on practice beyond classroom instruction.
• Collaborated on curriculum development to ensure courses remained current and pedagogically sound. Balanced technical rigor with approachability to support both beginners and advanced learners.

Backend Developer Intern at Prelax InfoTech

2021 - 2021

Developed and maintained RESTful APIs using Flask to support Android and iOS applications, enabling secure user authentication and fast, consistent data exchange; improved API response times by 30%.

• Optimized SQL queries and managed MySQL databases, achieving 40-50% reduction in query execution time while improving data retrieval performance and system scalability.
• Built and tested prototype e-commerce platform with recommendation engine using user behavior and purchase history to personalize suggestions, improving engagement by 25% in internal testing.
• Documented backend features, API specifications, and database schemas using Swagger and Markdown, improving onboarding speed for new developers and ensuring consistent communication with front-end and QA teams.
• Utilized comprehensive development toolkit including Postman for API testing, Git & GitHub for version control, and Docker for containerizing backend services for local testing and deployment simulation.
• Participated in agile development processes including code reviews, sprint planning, and bug tracking while collaborating cross-functionally with front-end and product teams.

Lecturer in Mathematics & IT at The SCIENCE Channel

2017 - 2021

Designed and delivered modern curriculum for Mathematics and IT subjects, integrating digital tools like GeoGebra and Desmos to enhance visualization, interactive learning, and student engagement.

• Introduced students to foundational computing concepts including logic gates, binary and number system conversions, programming fundamentals, and basic networking, providing early exposure to computer science principles.
• Taught essential problem-solving and algorithmic thinking, reinforcing skills that connect mathematics to real-world technology applications such as data processing, automation, and systems design.
• Mentored students for competitive entrance exams (GRE, Cambridge, CAT, JEE, CET, CPT), creating structured study plans and mock exams that led to successful outcomes in higher education admissions.
• Assessed student progress using diverse evaluation methods including assignments, projects, and oral reviews to ensure understanding and retention while providing individualized academic support.
• Contributed to educational innovation by participating in federal proposal drafting to advocate for technology integration in mathematics education and student-centered, adaptive teaching approaches.

Network Security System - MLOps Project

Built production-ready MLOps pipeline achieving automated threat detection for phishing URLs and malicious network traffic through end-to-end ML lifecycle management. Implemented modular pipeline architecture with real-time prediction API, automated data validation, and drift detection capabilities using Python, scikit-learn, FastAPI, and MLflow. Deployed scalable system on AWS with CI/CD automation via GitHub Actions, ECR containerization, and S3 storage, enabling automated model retraining and serving production traffic with experiment tracking and schema validation.

End-to-End FoodVision MLOps Pipeline

Developed a modular system using PyTorch and EfficientNet transfer learning, achieving 90%+ classification accuracy on pizza, steak, and sushi, with TensorBoard integrated for systematic experiment tracking. Architected and built two parallel web interfaces using both Flask and FastAPI, demonstrating framework flexibility and a clean separation of concerns. Implemented a non-blocking, asynchronous background training process with real-time status monitoring, ensuring a responsive UI and full cross-platform compatibility (CPU/CUDA/Apple Silicon).

AWS SageMaker Machine Learning Pipeline - Mobile Price Classification System

Engineered an end-to-end ML pipeline on AWS SageMaker, achieving 88% prediction accuracy for mobile price classification. Architected a cloud-native solution utilizing S3 for data management, IAM for security, and SageMaker for automated model training and deployment, showcasing MLOps best practices in a production-style workflow.

Text Summarizer Using HuggingFace Transformers

Achieved ROUGE-optimized summarization performance by developing production-ready text summarization system processing conversational data and meeting transcripts. Implemented end-to-end ML pipeline with HuggingFace Transformers (Pegasus model), data ingestion/transformation pipelines, and fine-tuning on SAMSum dataset via Google Colab GPU. Deployed RESTful API with FastAPI, Docker containerization, Weights & Biases experiment tracking, and comprehensive logging, delivering scalable ML service with automated pipeline stages and seamless deployment capabilities.

Student Performance Prediction System - End-to-End ML Engineering Project

Achieved 90%+ prediction accuracy by developing end-to-end ML web application predicting student math scores, bridging the gap between experimental ML models and production-ready systems. Architected modular Flask application with scikit-learn pipelines, comprehensive logging, and exception handling, deploying on AWS EC2 using Elastic Beanstalk with automated model selection from 7 algorithms. Delivered production-ready ML system demonstrating ML engineering, cloud deployment, and software architecture principles for data science and full-stack development applications.

Multi-Agent Financial AI System

Reduced manual research time by 95% by building multi-agent AI system using Python, Groq AI models, and Agno framework for automated stock analysis. Orchestrated specialized AI agents with Yahoo Finance API integration and web search capabilities, implementing agent coordination patterns and task distribution algorithms. Developed interactive Streamlit interface delivering real-time market data, analyst recommendations, and sentiment analysis with comprehensive financial insights and automated report generation.

Multi-Tier AI Agent System with Vector Database Integration

Engineered multi-tier AI agent architecture implementing three progressive complexity levels from simple web-search agents to coordinated multi-agent teams for financial analysis. Integrated multiple AI models (Groq, Gemini, OpenAI) with vector database (LanceDB) for knowledge management, hybrid search capabilities, and PDF knowledge bases. Demonstrated advanced agent coordination, domain-specific expertise, and scalable agent orchestration using Python, Agno framework, and DuckDuckGo/YFinance APIs.

Intelligent Document Q&A System with RAG Architecture

Delivered sub-second query response times by developing enterprise-grade RAG application enabling natural language querying of large PDF document collections. Implemented end-to-end document processing pipeline with vector embeddings, similarity search, and context-aware response generation using Groq API (Gemma model), Google Generative AI embeddings, and FAISS vector database. Built production-ready application with optimized chunking strategies, session management, and Streamlit frontend, demonstrating expertise in AI/ML engineering and scalable vector database architecture.

AI-Powered Blog Content Generator | AWS Serverless Architecture

Built production-ready serverless API leveraging AWS Bedrock's Meta Llama 3 for automated blog content generation with scalable cloud infrastructure. Architected end-to-end serverless solution integrating Lambda functions, API Gateway, and S3 storage with comprehensive IAM security policies. Implemented robust error handling, timeout management, and logging strategies for reliable cloud service orchestration, demonstrating expertise in serverless architecture patterns, AI model integration, and scalable infrastructure design.