I build offline LLM systems, multi-agent workflows, and cognitive layers that run on real hardware instead of living in slide decks. Most of my work sits between research and practical tools: local first assistants, RL tuned retrieval, belief style memory systems, and code tools that try to behave consistently under real usage.
As the founder of Aftermath Technologies, I take ideas from sketch to working prototype, then push on them until they are stable enough for other people to run without babysitting.
- Offline LLM and agent systems on consumer GPUs and Linux
- Multi-agent orchestration for code analysis, refactoring, and style enforcement
- RL driven retrieval and fusion across RAG, cache, and graph sources
- Belief ecology, contradiction tracing, and anomaly detection for long running sessions
- WordPress and web infrastructure that actually stays online and pays bills
Most repos fall into one of these buckets:
-
Cognitive layer experiments
Belief ecologies, tension, memory decay, conflict tracking, and other ways of modeling what a system thinks it knows. -
RL and orchestration
Local first orchestrators, agent swarms, and RL tuned fusion layers that sit in front of LLMs instead of treating them like a magic box. -
Code tools
Multi-agent code style enforcement, refactoring helpers, and utilities that try to keep human codebases from turning into sludge. -
Infra and web
Pieces that support Linux based hosting, WordPress, and the glue that keeps small businesses online.
Most of this is experimental but runnable. If something is rough, I try to at least document what it is supposed to do and how to start it.
Core work
- Python on Linux (Ubuntu)
- Offline LLM orchestration, multi-agent systems, reinforcement learning pipelines
- Retrieval and fusion across RAG, cache, and graphs, with evaluation harnesses for drift and failure modes
- Cognitive style layers: belief ecology, tension, contradiction tracing, long horizon sessions
AI and ML
- PyTorch
- LLM tooling and evaluation
- Prompt and agent design
- Anomaly detection and basic safety checks
- Offline RL for policies and fusion weights
Systems and infra
- FastAPI and async backends
- Redis, SQL and MySQL
- Docker, Git, basic observability and benchmarking
- Local first, privacy focused setups that do not depend on a cloud crutch
Web and WordPress
- WordPress, WooCommerce, custom themes and plugins
- PHP, JavaScript, HTML, CSS
- Performance tuning, migrations, security hardening, DNS and SSL
- Work on serious AI systems that have to be stable, explainable, and measurable
- Roles where Python on Linux, offline LLMs, RL, and multi-agent orchestration are core to the job
- Chances to turn the cognitive layer and belief ecology work into tools other people can run, inspect, and extend
- Email: bradkinnard@proton.me
- LinkedIn: https://www.linkedin.com/in/brad-kinnard/
- Website: https://aftermathtech.com
If you are working on real AI systems, cognitive layers, or infrastructure that needs to run without constant cloud support, feel free to reach out.
© 2025 Bradley R. Kinnard / Aftermath Technologies