Building multi-agent operating systems and applied AI workflows.
Working across the AI stack: inference economics, model behavior, context systems, agent harnesses, and adoption.
Portfolio · Projects · Research · Writing · Workshops
Multi-Agent Systems · OpenClaw · Context Engineering · Harness Engineering · Inference Economics
War Loops · Clawdbot · GTM Agentic OS · Frontier Cluster Lab
LinkedIn · X · Superagency Newsletter · Email
I am focused on agents and applied AI: how model capability becomes reliable work through context, harnesses, orchestration, evaluation, and product workflows.
The working map is a five-layer AI stack:
| Layer | What I am studying and building |
|---|---|
| Infrastructure & inference economics | Compute, APIs, model pricing, deployment constraints, and cost-aware systems |
| Models & learning | Model behaviour, efficient architectures, parameter-constrained experiments, and research translation |
| Context, data & memory | Context engineering, retrieval, memory, state, and workflow-specific knowledge |
| Agents & harnesses | Tool use, guardrails, multi-agent orchestration, harness engineering, and evaluation loops |
| Products & adoption | Applied AI workflows, enterprise adoption, AI-assisted work, and local community building |
Now
- Agents and applied AI systems
- OpenClaw, Clawdbot, War Loops, FlowState, and GTM Agentic OS
- Context engineering and harness engineering
- Understanding the five-layer AI stack across infrastructure, models, context, agents, and applications
- AI systems research through Parameter Golf and Token Index
Research & Systems Interests
- Context engineering and harness engineering
- Agent reliability, memory systems, and evaluation loops
- Inference economics: pricing, compute markets, country-level cost structure, and AI affordability
- Federated learning and distributed training
- Turning research ideas into durable tools
| Project | What it is | Status |
|---|---|---|
war_loops |
Autonomous frontend designer that captures URLs or images, extracts design specs, generates static and moving builds, and repairs them through judge-gated fidelity loops | 🟢 Active |
zzp_bunq |
FlowState: multi-agent financial ops for Dutch freelancers on live bunq APIs, with OpenClaw orchestration and CrabTrap guardrails | 🟢 Active |
gtm-agentic-os |
GTM Agentic OS for developer opportunity radar, OpenClaw workflow, technical proof building, and human-gated response packages | 🟢 Active |
prax |
Compute/API credits marketplace with Solana Anchor auctions, Token-2022 escrow, and a Next.js frontend | 🟢 Active |
clawdbot-the-endgame |
Local-first multi-agent orchestration OS for research, hiring, workflows, and structured execution | 🟢 Active |
| Project | What it is | Status |
|---|---|---|
parameter-golf |
OpenAI Parameter Golf experiments for parameter-constrained language modeling, ablations, and efficient architecture tradeoffs | 🟢 Active |
inference_economics |
Frontier Cluster Lab: interactive learning interface for decode latency, batching, KV cache, topology, utilization, and cost-aware AI serving | 🟢 Active |
token-index-global |
Sourced editorial model of coding-assistant inference spend across countries using model pricing, electricity, and price-level proxies | 🟢 Active |
sorachain_ai |
Prior whitepaper and prototype work on blockchain-native federated learning, model-to-data training, and privacy-preserving AI coordination | 📄 Research artifact |
| Project | What it is | Status |
|---|---|---|
openclaw_mirofish_outreach |
OpenClaw outreach workflow for context-aware prospect research, personalized messaging, and human-gated execution | 🟢 Active |
- System-design-first approach: architecture, constraints, and failure modes before implementation.
- AI-assisted build loop: Codex, Claude Code, and OpenClaw as force multipliers for speed and breadth.
- Agent framework experience across CrewAI, LangChain, and custom orchestration patterns.
- Agent reliability: context engineering, harness engineering, memory design, evaluation, and long-horizon orchestration.
- Applied AI systems that connect models, tools, workflows, and production adoption.
- Foundational AI research through parameter-constrained modeling and efficient architectures.
- Inference economics: model pricing, compute markets, affordability, and cost-aware AI deployment.
- Prior federated and privacy-aware AI research through SoraChain AI.
- 10+ years in distributed architecture, solutioning and technical strategy.
- Runs AI upskilling workshops through Build Superagency.
- Previously built SoraChain AI whitepaper and prototypes.
- Speaker at NVIDIA, Global Open Source AI Conference (GLOSAIC), and OpenAI community events.
- Collaborating on foundational AI research (model efficiency, agent reliability, evaluation systems)
- Building multi-agent systems for enterprises from the ground up with strong architecture and execution discipline
- Working with teams translating AI research into production-grade tools and workflows
- Technical talks and advisory