Narrative
The island is quiet.
Parts Remaining
8
John Swindell | CS 499 Capstone
When I first began my Computer Science program at SNHU, I operated under the misconception that the pinnacle of development was being a full-stack software engineer writing in languages like Haskell. However, exposure to a wide variety of programming disciplines radically shifted my perspective. I quickly realized my true passion lay in backend infrastructure, data pipelines, and Python. My coursework, combined with professional experience in financial machine learning, has driven my ambition to specialize in Platform Engineering and MLOps. This ePortfolio reflects that evolution, demonstrating my transition from writing isolated scripts to architecting scalable, cloud-native, AI-driven platforms.
The core of my capstone revolves around the total architectural overhaul of an early-degree Python scripting project. I completely decoupled the original monolithic application, containerizing the game logic with Docker and exposing it as a RESTful microservice using FastAPI. To showcase modern deployment practices, I utilized Terraform as an Infrastructure as Code (IaC) tool to programmatically provision serverless environments via Google Cloud Platform's Cloud Run.
Furthermore, I modernized the application's persistence layer by integrating MongoDB Atlas. This went beyond basic NoSQL document storage; I implemented Atlas Vector Search to construct a Retrieval-Augmented Generation (RAG) pipeline. By injecting semantic context into the Gemini API, I enabled dynamic, on-the-fly narrative generation.
Security is a foundational element of any platform engineering role, and I approached this architecture with a strict security mindset. I implemented Cross-Origin Resource Sharing (CORS) middleware to protect the API and utilized GCP Secret Manager to securely inject API keys at runtime, ensuring no credentials were ever committed to version control. This philosophy extends to my work outside the classroom. Professionally, I build secure ETL data pipelines for crypto assets. Personally, I manage a bare-metal home lab running a distributed cluster of local open-weight LLMs via Llama RPC, heavily utilizing Docker to sandbox AI agents and mitigate the risk of unauthorized system commands.
To push my technical boundaries, I replaced my capstone's static map dictionary with a dynamic procedural generation engine. I implemented a Binary Space Partitioning (BSP) algorithm to construct the dungeon rooms server-side. I then translated this layout into a Graph data structure to facilitate intelligent pathfinding.
By implementing the A* (A-Star) algorithm without diagonal movement, I engineered an NPC capable of actively hunting the player. Because my professional work heavily involves 2D NumPy matrices and Pandas DataFrames, conceptualizing spatial traversal through an adjacency list was a deliberate challenge. This implementation solidified my grasp of space and time complexity, proving my ability to select the most efficient data structures for demanding computational tasks.
Clear communication is essential when bridging the gap between software development and infrastructure operations. During Milestone 1, I conducted a formal Code Review video, analyzing legacy code, identifying bottlenecks, and successfully communicating a highly technical refactoring plan to stakeholders.
I apply this same communication strategy to safeguard intellectual property. When I wanted to showcase the complex crypto data pipelines I built for my employer, I navigated strict IP constraints by drafting detailed architectural diagrams and blog posts instead of sharing proprietary code. This allowed me to effectively communicate my system design expertise without compromising operational security.
The artifacts within this ePortfolio are not isolated assignments; they represent a cohesive, production-grade system. By synthesizing complex algorithmic logic (BSP/A*), robust database management (MongoDB/Vector Search), and automated cloud infrastructure (Terraform/Docker), I have demonstrated the comprehensive skill set required of a modern Platform Engineer. As I look toward the future, I am actively expanding my distributed hardware cluster and studying for the Certified Kubernetes Administrator (CKA) exam, ensuring my technical capabilities continue to scale alongside the rapidly evolving MLOps landscape.
Narrative