Asa Shepard

Researcher · Reliable AI agents & autonomous systems

I work on making AI agents reliable in the real world; closing the gap between benchmark performance and what actually holds up under production conditions. Currently focused on research into what moves the needle on agent reliability, and the infrastructure needed to support it.

Blog Writing on AI coding tools, context engineering, and autonomous agents →

Currently

Aspect Research Research on reliable AI agents (since Oct 2025)

aspectresearch.org

Research

Iterative Probe-and-Refine Tuning of Repository Guidance for AI Coding Agents Refining repo-level guidance for AI coding agents via synthetic probes; guidance helps on SWE-bench Verified (Jan-Apr 2026)

preprint

Structural Grammar of the Voynich Manuscript: Characterization and Mechanism Discrimination Exploring the Voynich manuscript through 40 hours of long-running agentic work (Apr 2026)

preprint

Recursive RAG Eval (World Wide Technology project) Evaluation framework for recursive exploration vs retrieval-augmented generation (Jan 2026)

github

Hackathons

TreeHacks 2026: Santé Multi-agent speech analysis for biomarkers (Feb 2026)

devpost

Other Projects

Campus Web App Williams College Marriage Pact analysis (2025–Jan 2026)

williamstf.com (williams.edu only)

FocalRouter Intelligent request routing for LLM APIs (2025)

github

HeatSheet F# DSL for optimizing track meet lineups (2025)

github

Minecraft Plugins Suite of ~10 Java plugins (~2021–2022); first shipped code — top plugin at 7,000 downloads

dev.bukkit.org

Previously

Yosa Co-founder; HIPAA-aligned mental wellness platform with 30 active providers (2025–2026)

yosaapp.com

OnCabaret Contract software engineer; social app for a NYC startup with 2,000 users (2025)

oncabaret.com

Web Surfing Studios Software engineer intern; built full-stack web applications (2024)

websurfingstudios.com

WSO (Williams Students Online) Front-end developer and tester (2024)

wso.williams.edu

theme: dark