Leslie Osei-Anane

Builder, researcher, soccer player. Studying Computer Science at The College of Wooster.

Currently: software engineering intern at The College of Wooster (Summer 2026), researching agent reliability, and studying CS at The College of Wooster.

Previously: SOREP student researcher, teaching assistant at The College of Wooster, AI/ML scholar at Amazon, and participant at MHacks 2025 / NexHacks 2026.

Projects

Holodeck - Generative VR world engine that turns room descriptions into schema-checked VR environments with LLM planning, asset ranking, and Meta Quest delivery.
Lumon - Real-time browser-agent supervision with frame streaming, sprite state, target markers, approval prompts, and artifact-backed review.
Roulette - Generative UI system that serves random one-off interfaces, games, tools, and simulators with queued generation, sandboxing, deduplication, and quality checks.
Conway Pattern Evolution Engine - Genetic algorithm for evolving long-lived Conway's Game of Life patterns.
RL-Connect-4 - Playable Connect Four web app with human, self-play, and semi-random RL modes.
Toolsmith - Cloudflare Workers AI agent platform that turns OpenAPI specifications into executable AI skills.
Handoff - MemoryLayer-powered app for importing AI chat history and generating portable context blocks.
Ghost - Privacy-first macOS assistant with screen-aware memory, voice control, semantic recall, and explainable memory graphs.
HiLite - Django app that extracts highlighted text from PDFs and exports annotations to Notion.

Experience

Algoverse AI Research, AI Research Intern - built Failure Workflow Memory, reducing post-policy failure rates by up to 34 percentage points across 4 frontier models and a 250-sample ToolBench evaluation set.
College of Wooster, incoming Software Engineering Intern - SOREP, Summer 2026.
College of Wooster, Student Researcher - SOREP - developed genome annotation clustering and search pipelines across 74,000 sequences and 18,000+ alignments.
Amazon AI/ML Scholar - built InterviewPrep.AI with prompt chaining and AWS SageMaker pipelines.
College of Wooster, Teaching Assistant.