Projects & Work

SouthDenverPlumber.com is a modern, responsive website I built for Frank Santilli's plumbing business serving the South Denver metro area, including Highlands Ranch, Littleton, Lakewood, Centennial, and Lone Tree. The Business Challenge Frank needed a professional online presence that would: Clearly showcase his plumbing services to potential customers Make it easy for clients to contact him for emergencies or bookings Build trust with potential customers through testimonials and service guarantees Rank well in local search results for the South Denver area What I Built I developed a user-friendly website with a clean, professional design that reflects the reliability of Frank's services. The site includes: Comprehensive Service PagesDetailed descriptions of all services including water heater repair, drain cleaning, residential and commercial plumbing services, emergency services, and preventative maintenance. Optimized Contact FlowProminent display of phone numbers and contact forms, making it simple for customers to reach out during plumbing emergencies. Customer TestimonialsA showcase of real customer experiences to build trust with potential clients. Service Area HighlightingClear information about the neighborhoods and areas Frank serves in South Denver. Mobile-First DesignFully responsive interface that works perfectly on phones, tablets, and desktops - essential for customers who often need to find plumbing help while away from their computer. Technical Implementation The website was built using: React with TypeScript for a maintainable, type-safe codebase Vite for fast build times and optimal performance TailwindCSS for responsive, custom styling Framer Motion for subtle animations that enhance the user experience React Router for client-side navigation Custom SEO optimization for local search performance Business Impact The new website has helped Frank: Increase local visibility for his plumbing business Generate more direct inquiries through the website Streamline customer communication Present a professional image that stands out from competitors Build a foundation for future digital marketing efforts Visit the Site You can see the live site at SouthDenverPlumber.com This project demonstrates how even traditional service businesses can benefit from modern web development techniques, delivering better customer experiences and business outcomes.

If you've ever wished you could easily reconnect with everyone you've met through your calendar—a10d.info is a tool I built to make this dead simple. This lightweight tool gives you the ability to upload your .ics calendar file and quickly extract a filtered list of meeting attendees—complete with their email addresses. Why I Built This There were so many times I wanted to: Follow up with people I’d met in recent months. Create a small email update list without exporting my entire address book. Audit how I was spending my time and with whom. Google Calendar doesn’t make this easy out of the box. So I built something that does. What You Can Do with a10d.info Once you upload your .ics file, you can: Filter your resultsChoose your desired date range, cap the maximum number of attendees per meeting (to avoid pulling emails from giant webinars), and optionally limit to meetings you actually accepted. Export attendeesThe tool compiles the filtered list of attendee emails across all matching meetings. You can export this to a CSV and use it for: A lightweight mailing list Re-engagement campaigns Rebuilding your network from past activity See your statsYou'll get an analytics dashboard with insights like: Total meetings during your selected time period Unique attendees Average attendees per meeting Events you organized vs. attended Your longest meeting Most common meeting day Meeting frequency over time Here’s a sample stats dashboard from my own data: Built With This project is powered by: TypeScript + Vite for the frontend ical.js for parsing calendar data Vercel for fast, free hosting A simple, clean interface that stays out of your way Try It Out The site is free to use and doesn’t store your calendar data—everything runs client-side during the file upload and processing phase. 👉 Visit a10d.info Whether you're job hunting, consulting, doing outreach, or just curious about your own meeting patterns—this little tool can help you turn passive calendar data into something useful. Let me know what you think—or shoot me feature ideas if there's something you'd love to see added.

I am using Python to build a system that will manage events around the world for a community building organization. It will allow: Users to sign up, refer friends, manage their profiles Admins to manage the locations of groups, the local managers, the events, and users Local Managers will be able to manage their location, their events, sponsors, and users Create Slack channels on demand as new groups or locations are created This is to replace multiple spreadsheets, siloed information, and provide more organization for a large scale world-wide community building group.

AI Text Processing Pipeline Overview I'm developing an intelligent text processing solution that automates the extraction of valuable data from thousands of unstructured text files in the high-end luxury product market. Rather than crafting and maintaining custom parsers for each data source—a process that becomes unwieldy as sources evolve and formats vary—I've engineered a sophisticated prompt system that harnesses the power of ChatGPT to transform raw text into structured JSON data. This approach offers remarkable adaptability in the face of inconsistent data sources and evolving formats, reducing both development overhead and maintenance complexity. By delegating the pattern recognition and extraction tasks to an LLM through carefully calibrated prompts, the system can identify and extract relevant information with minimal human intervention, saving days of work per data source. Technologies Used Python for application architecture and data orchestration ChatGPT API for natural language understanding and data extraction JSON for structured data representation Advanced caching strategies for parallel processing and API rate limit management Workflow pipeline architecture (similar to Temporal.io) for reliable processing Database technologies for persistent storage and retrieval Custom prompt engineering for reliable extraction patterns My Role As the architect and lead developer of this project, I am: Designing the end-to-end data processing architecture Implementing robust error handling and fallback mechanisms Developing and refining the prompt engineering to ensure consistent extraction Creating validation systems to verify data quality and completeness Building scalable processes to handle growing volumes of text Optimizing parallel processing workflows with intelligent caching Challenges and Solutions Working with AI-powered text extraction presents several interesting challenges: Prompt Engineering: Crafting precise instructions that consistently yield the correct data structure across varied inputs Model Variation Handling: Building resilience against API and model changes over time Inconsistent Source Data: Implementing adaptive approaches to handle missing fields and format inconsistencies Validation Mechanisms: Cross-referencing data across multiple sources to ensure accuracy Scale Processing: Managing the extraction of data from 30+ distinct sources efficiently Rate Limiting: Implementing sophisticated caching to optimize API usage while maintaining throughput Project Scale The initial phase of this project involves processing: Thousands of unique luxury products Approximately 20,000 product images Roughly 2,500 product listings Data from more than 30 distinct sources The system is designed for continuous expansion, with plans to incorporate additional data sources and enhance existing product profiles over time. Outcomes As an ongoing project (initiated in March 2025), the current achievements include: A functional data ingestion and processing pipeline Reliable JSON transformation of unstructured text Comprehensive product profiles that aggregate information across multiple sources Scalable database architecture for the processed data Next Milestones Once the data extraction and structuring pipeline reaches production stability: Develop an intuitive front-end interface to visualize comprehensive product profiles Create systems to track product history, including previous listings and ownership changes Implement comprehensive user management with authentication Deploy flexible monetization options including subscription models and single-purchase access Add engagement features such as referral systems and promotional discounting

AI Image Recognition Project Overview I am working on building a software solution that utilizes the EfficientNet-B3 Model, a state-of-the-art convolutional neural network for image recognition. The project focuses on training the model to recognize specific patterns within images to match them with product listings found online. This application has significant potential for e-commerce, inventory management, and competitive analysis use cases. Technologies Used Python for core development and model implementation EfficientNet-B3 neural network architecture Machine Learning frameworks (TensorFlow/PyTorch) Image analysis and computer vision techniques Data processing and augmentation pipelines My Role As the creator and developer of this project at Bernier LLC, I am: Designing and implementing the core machine learning architecture Training the EfficientNet-B3 Model on custom datasets Developing pattern recognition algorithms for product matching Building the end-to-end pipeline from image input to product listing matches Testing and optimizing the model for accuracy and performance Challenges and Solutions Working with image recognition for product matching presents several unique challenges: Variations in product photography styles and quality Handling different angles, lighting conditions, and backgrounds Ensuring high accuracy to make the matches commercially viable Optimizing the model to run efficiently with reasonable computational resources To address these challenges, I'm implementing extensive data augmentation techniques to improve model robustness, using transfer learning to leverage pre-trained weights, and developing a custom similarity scoring system to improve match confidence. Outcomes As this is an ongoing project (started March 2025), the key outcomes so far include: Successful proof of concept demonstrating pattern recognition capabilities Development of a training pipeline for the EfficientNet-B3 model Initial testing showing promising results for product matching accuracy Framework for integrating with e-commerce platforms for real-world application The project continues to evolve as I refine the model and expand its capabilities for commercial applications.

Software Analysis Tool Overview This project is developing a specialized analysis tool that examines GitHub repositories to uncover insights and information about codebases that might not be immediately obvious to developers. The tool integrates AI capabilities to enhance the analysis process and provide intelligent feedback on code architecture, potential issues, and improvement opportunities. By leveraging both traditional static analysis techniques and modern AI approaches, the tool aims to give developers a deeper understanding of their codebase. Technologies Used Python as the primary programming language GitHub API for repository access and integration AI/ML frameworks for code pattern recognition Static code analysis techniques Natural language processing for documentation analysis My Role As the creator and lead developer at Bernier LLC, I am: Designing the overall architecture of the analysis system Developing the core analysis algorithms and metrics Implementing the GitHub integration components Building and fine-tuning the AI analysis capabilities Creating the reporting interfaces and visualization tools Challenges and Solutions Developing a sophisticated code analysis tool presents several significant challenges: Processing and analyzing repositories of varying sizes and structures Identifying non-obvious patterns and insights beyond what traditional tools offer Integrating AI capabilities in a way that produces actionable insights Balancing depth of analysis with performance considerations I'm addressing these challenges by implementing a modular architecture that allows for specialized analyzers for different types of code and repositories, developing custom heuristics to identify code patterns, and carefully integrating AI components that can be trained on specific types of code structures. Outcomes As an ongoing project (started November 2024), the tool has already achieved: Successful integration with GitHub's API for repository analysis Implementation of baseline static analysis capabilities Development of initial AI-powered insights for Python codebases Framework for extensible analysis plugins to support various programming languages The project is continuing to evolve with plans to add more programming language support, deeper AI integration, and enhanced visualization capabilities to make complex code relationships more accessible to developers.

Created a project management dashboard that consolidated data from four different task management systems, providing leadership with a centralized view of priorities, progress, and blockers. Built using Python, APIs, and visualization tools, the system streamlined project tracking and improved cross-team alignment.

Initiated and led the Product-Led Growth (PLG) initiative at Temporal Technologies, aligning over 30 stakeholders to improve self-signup experiences. Achieved a 14% conversion rate within three months, leading to 1 billion user actions within six months. Integrated PLG strategies with customer onboarding to accelerate adoption.

Built and managed an e-commerce business focused on selling high-quality barn doors to interior designers, contractors, and homeowners. Oversaw website design, marketing, order fulfillment, and customer support, growing the business through targeted outreach and customer-centric service.