Ryan's Portfolio

My passion for AI started long before the current hype cycle. I recognized the potential of Recurrent Neural Networks (RNNs) early on. For example, at Graphic Products, I built a solution using word vectors to automate the extraction of accident statistics from hundreds of OSHA records, turning a manual research nightmare into an instant resource.

At Precoa, I tackled a critical quality assurance issue where broken data pipelines were sending flawed reports to customers. I architected a "Report Checking AI" using AWS EventBridge and Lambda to orchestrate validation. The system pulls report images via the Tableau API and sends them to OpenAI for verification against human-readable rules. To ensure accuracy, I implemented a consensus mechanism, querying the model three times and using the majority vote, which drastically reduced false positives. If a report fails, the system automatically pauses the subscription and alerts the team via SMS before the customer ever sees it.

I also apply these tools to everyday problems. I recently mentored a friend in the auto body industry, teaching him to build a custom GPT that processes shipping labels and verifies part numbers. This simple tool saved him 3-4 hours of manual data entry every single day.