I am a Software Engineer and Graduate Student pursuing a Master of Science in Artificial Intelligence Systems at the University of Florida (Aug 2025 – Jun 2027). With 3+ years of professional experience at Accenture and Temenos, I am passionate about designing scalable, intelligent systems that combine cloud infrastructure, AI, and automation to drive business transformation.

• 🔭 Currently working on AI-powered applications and ML infrastructure
• 🌱 Learning advanced deep learning, generative AI, and MLOps
• 💡 Interested in Responsible AI, Generative AI, Model Deployment, and Cloud-Native AI Solutions
• 📫 Reach me at: harshal27patel@gmail.com
• 📍 Based in Gainesville, Florida, USA

Dec 2024 – Aug 2025 | Bengaluru, India

• Built scalable web applications and microservices processing 10M+ records daily with 99.9% uptime
• Developed RESTful APIs and data pipelines with Python and Docker, reducing API response time by 40%
• Architected ML infrastructure automation using AWS, reducing model deployment latency by 30%
• Optimized CI/CD pipelines with GitHub Actions, enabling teams to ship production features 2x faster

Sep 2021 – Dec 2024 | Bengaluru, India

• Shipped backend services for banking applications handling 100K+ transactions daily
• Designed REST APIs with Redis caching, achieving 60% reduction in query response time
• Refactored monolithic architecture into microservices, improving scalability to handle 3x peak traffic
• Built automated testing framework with 85%+ code coverage, reducing production bugs by 45%
• Applied ML for anomaly detection in banking systems, improving fraud detection by 25%

Tech Stack: Python, PyTorch, Hugging Face Transformers (SigLIP 2, DistilBERT), FastAPI, Docker

• Architecting a production-grade multimodal fraud detection pipeline fusing computer vision (SigLIP 2), NLP (DistilBERT), and structured data analysis
• Designing a cross-modal attention fusion layer to dynamically weigh modalities and produce a unified fraud score
• Integrating explainability using Captum to provide actionable insights via image heatmaps and text token attributions

Tech Stack: Django 4.2, Django REST Framework, React 18, PostgreSQL, Anthropic Claude, Docker

A full-stack support ticket management system with AI-powered ticket classification using LLMs.

• Built end-to-end ticket management with create, read, update, filter by category/priority/status, and search
• Implemented real-time statistics dashboard with database-level aggregations for optimal performance
• Developed hybrid AI classification using Anthropic Claude Sonnet 4 with intelligent keyword-based fallback
• System auto-classifies tickets by category (Technical, Billing, Account, General) and priority (Critical, High, Medium, Low) as users type
• Users can accept or override LLM suggestions, ensuring human-in-the-loop control
• Containerized with Docker Compose for one-command deployment with PostgreSQL and full-stack services

Tech Stack: Python, AIF360, LIME, SHAP, Streamlit, Docker, Prometheus

• Developed ML fairness auditing platform evaluating 15+ metrics with 92% accuracy
• Built production web application serving 1,000+ users with real-time data visualization
• Implemented monitoring infrastructure with Prometheus and Grafana, processing 10K+ analysis requests
• Designed automated compliance reporting with PDF generation and Slack alerts

Tech Stack: Python, Google Gemini 2.0, Flask, OpenCV, U2Net, Streamlit

• Engineered multimodal AI application enabling users to visualize 78,000+ fashion items
• Implemented computer vision pipeline with U2Net segmentation achieving 95% garment detection accuracy
• Processing 1,000+ user requests with sub-3-second response time
• Implemented Redis caching layer reducing API costs by 40%

Tech Stack: Python, n8n, OpenAI API, DALL-E, LinkedIn API

• Architected automated content generation system generating and scheduling 120+ posts monthly
• Built end-to-end automation pipeline with error handling and retry mechanisms
• Reduced content creation time by 90% while maintaining 99% successful publish rate

Tech Stack: Python, Scikit-learn, Ensemble Methods, Feature Engineering

• Built supervised learning pipeline achieving 88.1% accuracy on 500+ matches
• Published research paper at International Conference on Intelligent Computing (ICIIC-2021)
• Implemented SVM, Random Forest, and Logistic Regression with hyperparameter optimization

• 🔹 Deep Learning Specialization (NVIDIA)
• 🔹 AWS Cloud Technical Essentials
• 🔹 Introduction to Kubernetes
• 🔹 Azure AI Fundamentals
• 🔹 Google Cloud Platform Fundamentals: Core Infrastructure
• 🔹 Python for Data Science, AI & Development

"Prediction of IPL Match Outcome Using Machine Learning" International Conference on Intelligent Computing (ICIIC-2021) Comparative analysis of supervised learning algorithms achieving 88.1% accuracy with feature selection and ensemble methods.

• 🤖 Responsible AI & Model Deployment
• ☁️ Cloud-Native AI Solutions
• 🔧 AI Infrastructure Automation
• 🧠 Deep Learning & Neural Networks
• 🎨 Generative AI
• 📊 MLOps & Model Monitoring

I'm always open to interesting conversations and collaboration opportunities. Feel free to reach out!