Alan Rocque
About

Alan Rocque

Senior at the University of Wisconsin–Madison studying Computer Science and Data Science, graduating in May 2026.

I’ve spent the last three summers as a software engineering intern at Amazon and Intel. Most of what I build now sits in one place: tooling and evaluation for the next layer of AI systems. Memory services for agents, benchmarks for world models, scaffolding for MCP servers, reproductions of new research papers.

I like the part of the stack where systems engineering meets ML, where a clean interface matters as much as the model behind it.

alanrocque@outlook.com GitHub LinkedIn Resume
Selected projects

What I’ve been building

Four projects that fit the same thesis: infrastructure and evaluation for more capable AI systems.

Agent memory Open repo

Engram

Experiential memory service for AI agents. Ingests OpenTelemetry traces, distills reusable lessons with an LLM, and serves them at decision time via hybrid retrieval. FastAPI, Postgres with pgvector, Celery, React, a Python SDK, and end-to-end integrations.

Research reproduction Open repo

KimiHarness

Open-weight reproduction of Meta-Harness, built three weeks after the paper dropped. Replaced the proposer model with Kimi K2.6 via Moonshot and pushed Symptom2Disease accuracy from 0.69 to 1.00 over 12 iterations for 73 cents.

World models Open repo

DreamBench

Diagnostic benchmark for learned world models including DreamerV3, IRIS, DIAMOND, and Delta-IRIS. Runs targeted probes for object permanence, physics consistency, and reward fidelity across Atari, MiniGrid, and Crafter.

Developer tooling Open repo

create-mcp

npx create-mcp@latest scaffolds a production-ready Model Context Protocol server in TypeScript or Python with auth, deployment configuration, and passing tests.

Experience

Where I’ve worked

Summer 2025

Amazon

Software Engineering Intern — distributed configuration system on AWS AppConfig, including multi-region A/B testing across NA, EU, and FE.

Summers 2023 and 2024

Intel

Software Engineering Intern — anomaly detection for hardware log files and migration work on an internal workflow API from Node.js to NestJS.

2024 to present

UW Informatics Skunkworks

Undergraduate researcher — NBA game outcome prediction with SVMs and multi-layer perceptrons.