Oliver Clarke

Full-stack web developer working across two client SaaS products. Designed and built a LangChain-based RAG pipeline to ground LLM integrations in client-specific data, cutting token usage by up to 70% through tighter context retrieval and prompt design. Built full-stack components in Svelte and React, profiling and restructuring data loading to cut page load times by up to 40%. Built and integrated a complete Stripe billing system covering subscriptions, webhooks, plan changes, failed payments, and prorations. Also shipped a voice-to-text feature that improved feedback capture rates by roughly 50%.

Spent over three years building VR and simulation software with Propel Virtual Reality. Designed reusable architectural systems in Unity/C# that were shared across multiple client projects rather than rebuilt each time, working alongside a team of 5 engineers. Delivered training and simulation experiences for CAMH, Queen's University, uOttawa, and Lockheed Martin, each with distinct hardware, interaction, and use-case constraints.

Let's Cook is a full-stack kitchen companion that turns whatever is in your pantry into concrete meal ideas. Users track ingredients and expiry dates, and the app highlights what needs to be used first. A single tap sends your pantry, portions, and dietary prefs to the OpenAI API to generate structured recipes with steps, ingredients, and cook time, backed by Spoonacular for ingredient data and Supabase for auth and real-time sync. On the frontend, React Native components are organized around a single pantry state so recipe generation, expiry alerts, and dietary filters all stay in sync.

Project Draftr is a full-stack analysis tool for five-champion League of Legends team comps. It pulls live win rates by rank from the Riot Games API, scores pairwise champion synergy, and classifies each draft into an archetype (e.g. teamfight, pick, front-to-back) with a letter grade so players can quickly understand strengths, gaps, and win conditions. Under the hood, a FastAPI backend handles data fetching, caching, and composition scoring, exposed to a React frontend that focuses on clear visual feedback as you tweak drafts.

SPECTRE QUEST began as a small passion project and grew into a large-scale Unity title built over four years. It pushed my skills in building reusable gameplay systems, optimizing performance for complex scenes, and putting real project management in place around scope, milestones, and collaboration. The codebase is structured around modular systems (abilities, enemies, levels) that can be extended without rewriting core game loops, which was critical as scope evolved.