Builder and founder at the intersection of AI agents, finance infrastructure, and shipping real products.
I build and maintain large-scale systems as cohesive ecosystems, with a strong emphasis on correctness, safety, and production readiness.
Audit infrastructure for AI agents: a verifiable record of what agents did.
- Append-only audit event log
- Reliability signals per audit scenario
- Cryptographic attestation (HMAC-SHA256)
- Human review queues
- Tenant isolation with RLS
- Python/TypeScript SDKs, MCP server, public verification
🌐 Sigmodx · 🔗 Sigmodx-Public · 📦 sdk-python · sdk-typescript · mcp-server · integrations-python
A Python-native multi-agent trading framework with a modular pipeline and safety-first architecture.
- Pipeline: regime → sentiment → insider → risk → allocation
- Hard circuit breakers and safety-first design
- Backtesting engine
- FastAPI orchestrator
- MCP interface
Agent memory optimized for what's ahead: plan-aware, forward-looking context compression for LLM agents.
- Compresses context around future plans, not just past history
- Python, Apache 2.0, open source
🔗 memahead/memahead · 🏢 memahead org
Economic control plane for AI agent fleets: enforced, delegated authority over what agents can spend.
- Agent-managed wallets with fleet budgets and capital caps
- Scoped MCP keys and append-only audit logs
- Human approval queues for policy-gated transactions
- Lending between agents and shared expense coordination
🌐 EmbiPay · 🔗 EmbiPay-public · 📦 EmbiPay-SDK · 🤖 Reference Agent
A live, cross-platform travel social network integrating real-time flight intelligence with AI-driven social experiences.
- 🤖 AI-powered recommendations and assistants
✈️ Real-time flight intelligence- 💬 Secure social interactions between verified travelers
- 📍 Airport discovery and time-aware planning
🔗 Layover AI: Public architecture & system design
🌐 Website · 🍎 App Store · 🤖 Google Play
Core source code is private. The public repository documents the system architecture, technical tradeoffs, and product decisions behind a shipped production platform.
Open-source execution intelligence for AI agents, an execution shadow that records what agents actually do.
- Framework-neutral, structured event protocol with causal links
- Python SDK, FastAPI collector, JSONL event store
- One-run viewer with execution graph, node list, and timeline
- Apache 2.0, built in the open
🔗 Tselora/Tselora · 🏢 Tselora org
I am actively building production-oriented open-source patterns for MCP (Model Context Protocol), focusing on safety, observability, testing, and deployment of agentic systems.
- mcp-server-template-python: Production-ready MCP server template (Python)
- mcp-toolkit-examples: Safe and scoped MCP tool design patterns
- mcp-agent-safety-playbook: Guidelines and best practices for safely running agentic AI with MCP tools in production
- mcp-deployment-patterns: Deployment patterns and operational best practices for running MCP servers safely in production
- mcp-observability: Observability patterns and best practices for monitoring MCP servers and agent tool usage
- mcp-testing-playbook: Testing and validation strategies for MCP servers and agent-exposed tools in production
AI / LLMs
Multi-agent systems, MCP, LangGraph, Pydantic AI, RAG, agent orchestration
Backend & Systems
Python, C# (.NET Core), FastAPI, PostgreSQL, TimescaleDB
Product & Interfaces
React Native, Streamlit, full-stack system design for real-world platforms
Designing agent-safe, observable, and testable MCP-based systems that integrate legacy infrastructure with modern AI agents, with a strong emphasis on production constraints and risk management.