Senior AI Engineer at Exodus, building AI-assisted developer tooling, agent workflows, and secure automation for engineering teams.

I focus on practical AI infrastructure: local-first tools, coding-agent observability, MCP/A2A workflows, LLM integrations, and privacy-conscious developer experience.

Local-first agent-forensics workbench for Codex and Claude Code JSONL sessions.

Perlustron helps developers understand what happened in an autonomous coding run: where it drifted, what failed, what looped, what changed, and what can be safely shared. It includes an interactive local UI, timeline/transcript views, parser-health diagnostics, redacted exports, diffing, insight queues, release archives, checksum manifests, CI, and privacy/security documentation.

Why it matters:

• helps maintainers debug and review coding-agent sessions without uploading private logs
• supports Codex and Claude Code session formats with schema-drift visibility
• emphasizes local execution, redaction, no telemetry, and safe sharing workflows
• is useful for OSS maintainers adopting coding agents for triage, review, release, and incident workflows

OpenAI-compatible bridge that lets Honcho use ChatGPT/Codex subscription models through a local /v1/chat/completions-style endpoint. It reuses Codex/Hermes ChatGPT OAuth credentials and avoids patching Honcho itself.

Profile-backed specialist-agent plugin for Hermes Agent, enabling a main Hermes assistant to call bounded specialist profiles with their own memory, skills, config, and model settings.

• Coding-agent observability and session forensics
• Local/private AI developer tools
• OpenAI-compatible bridges for agent infrastructure
• Hermes Agent plugins and multi-agent workflows
• Secure, reviewable automation for engineering teams

Exodus — Senior AI Engineer

• Building AI-powered developer tools for a fully remote engineering organization
• Driving developer productivity, automation, platform reliability, and AI workflow adoption
• Reviewing, hardening, and operationalizing AI-assisted engineering workflows

OpenText / AppRiver — AI, ML, data, and threat research engineering

• Built GenAI/LLM, RAG, threat-detection, ETL, and large-scale analysis systems
• Partnered across engineering, product, and security teams to move prototypes into production

Languages: Python, Rust, Go, Scala, TypeScript, JavaScript, C++, Java, C#

AI/agents: LLMs, coding agents, RAG, prompt engineering, MCP, A2A, Hermes Agent, Codex, Claude Code

Data/infrastructure: Databricks, Snowflake, Elastic, BigQuery, graph databases, Airflow, Spark, CI/CD, containers

• Email: bear@bearhuddleston.dev
• LinkedIn: linkedin.com/in/bearhuddleston
• Work GitHub: @Bewarden

Always happy to talk about AI developer tooling, agent workflows, and local-first OSS.