Software Engineer | Full-Stack | Cloud & DevOps | AI Systems

BS Software Engineering @ RIT (GPA 3.83)

Linde · May 2026 – Present

• Collaborating with a data scientist to design and deploy a full-stack production forecasting tool that automates volume planning for plant managers, replacing a manual process
• Built with ASP.NET and React TypeScript frontend, Postgres backend, deployed on Azure, Owning the full-stack implementation end to end

Accessible Learning Labs · Jan 2025 – Present

• Lead a team of 3 software engineers using Agile methodologies to design and implement educational labs on AI bias and hallucinations
• Architected and developed full-stack features using React, Node.js, and PostgreSQL for deployment across 9 academic institutions
• Built a research web platform (React, Node.js, PostgreSQL) used by 600+ participants and adopted for recurring studies
• Presented at CCNSE on the impact of accessible software in education and research

Rochester Regional Health · Aug 2025 – Jan 2026

• Implemented CI/CD automation (GitHub Actions, RenovateBot, NUnit), cutting maintenance effort by 15–20%
• Refactored and optimized 30+ .NET services, reducing backend execution time by 60%+ on critical paths
• Consistently delivered sprint work 20–40% ahead of estimates

React (TS) · FastAPI · AWS · Terraform · DynamoDB · SES · EC2 · xAI Grok · Twilio Built a document-aware AI pipeline that systematically parses, chunks, and indexes SEC 10-Ks to prevent LLM context loss.

• Streaming conversational model handles natural speech in real time while a parallel background model extracts structured appointment data (name, address, job type, date) to DynamoDB
• Per-turn DynamoDB writes + xAI store_messages + previous_response_id allow conversations to resume seamlessly after dropped WebSockets
• All client-specific behavior (prompts, voices, calendar IDs, greetings) lives in configuration — the core WebSocket handler stays static across clients
• Deployed on EC2 behind nginx (SSL via Certbot) with systemd for persistence and Terraform + S3-backed remote state for IaC

React (TS) · Python · AWS · Terraform · DynamoDB · S3
Built a document-aware AI pipeline that systematically parses, chunks, and indexes SEC 10-Ks to prevent LLM context loss.

• AI-generated summaries are created per section and cached in S3 for reuse
• Subsequent queries refetch pre-processed data instead of re-parsing raw filings
• Improves accuracy, latency, and cost vs naïve “upload PDF to LLM” approaches

React Native · Expo · Python · AWS · Terraform · PostgreSQL · Docker
Mobile app built for a track coach to collect previously unrecorded athlete workout data.

• Replaced email and paper logs with structured in-app submissions
• Centralized athlete times and metadata into a single coach dashboard
• Increased recorded practice data by ~600%. Grew to 20 DAUs

React (TS) · C# ASP.NET · AWS EC2 · Docker
End-to-end habit tracking web app focused on consistency and long-term progress.

• Users can create, edit, and delete habits with streak and historical progress tracking
• Social features for adding friends and comparing accountability
• AI-generated habit recommendations based on user behavior
• Containerized and deployed on EC2, owning backend, frontend, and infrastructure

Languages
Python, C#, Java, HTML, CSS, JavaScript, TypeScript, JavaScript

Frameworks
React, React Native, Expo, Flask, ASP.NET

Cloud / Infra / Data
Linux, SQL, Docker, MongoDB, Terraform, Github Actions, Postgres, AWS (EC2, Lambda, IAM, Amplify, Cloudfront, S3, SES, DynamoDB, RDS, Bedrock, Route53, RDS, Elastic Beanstalk, API Gateway)

• LinkedIn: linkedin.com/in/conner-jack-defeo
• Portfolio: connerdefeo.com