Mohsen Jahanshahi
I build systems that
run themselves.
Autonomous agent infrastructure, DeFi arbitrage, and local AI compute — deployed across a home lab running 24/7 on Proxmox, Docker, and NVIDIA GPUs.
Technical Founder with a home lab problem.
I run a Proxmox cluster in my apartment with GPU compute, a dozen Docker services, and AI agents that handle everything from trading alerts to infrastructure monitoring. Before tech, I managed production lines — the obsession with repeatable systems carried over.
Most of what I build falls into three buckets: agent infrastructure (semantic memory, multi-channel harnesses, self-growing skill registries), DeFi tooling (arbitrage detection, delta-neutral strategies on Solana), and B2B products (RAG pipelines for engineering firms, workflow automation).
I work primarily in TypeScript and Python, with Rust and Go for performance-critical pieces. Everything self-hosted where possible — Ollama for inference, ChromaDB for vectors, Cloudflare Tunnels for ingress.
Production Services
Running on local infra
Revenue Systems
Arb, SaaS, B2B
Agent Frameworks
ClawBox · Hairy · Madi
agentssot-apiUp 14d
rag-backendUp 14d
gpu-inferenceUp 7d
chromadbUp 14d
profit-machineUp 9d
vectordata-webUp 3d
$
What I'm Running
Selected from 40+ repositories
hari-hive
Semantic Memory for AI Agents
Persistent knowledge graph that gives any AI agent long-term memory. Stores facts, skills, and decisions as embeddings in ChromaDB, then surfaces them via semantic recall. Every agent I run — coding assistants, trading bots, infrastructure monitors — writes to and reads from this single source of truth. Cross-session context, skill teaching, and feedback loops that improve retrieval over time.
Polymarket Profit Machine
Delta-Neutral Arbitrage on Solana
Five arbitrage detection strategies scanning prediction markets for mispriced contracts. Opens hedged positions through Jupiter perpetual contracts to capture funding rate differentials while staying delta-neutral. Configurable thresholds, automated alerts via Telegram, and a dashboard for tracking P&L across strategies. Runs continuously on the home lab.
VectorData.solutions
Agentic AI for Engineering Firms
B2B platform that ingests engineering documents (specs, BOMs, datasheets) into a RAG pipeline and exposes an agentic parts-lookup interface. High-accuracy sourcing across ODB device catalogs. Team project with active clients.
Hairy
Multi-Channel Agent Harness
Standalone agent runtime with Telegram, WhatsApp, webhooks, and CLI channels. Self-growing skill registry with lifecycle promotion (draft → candidate → active → retired). Prompt versioning, a proactivity engine, and JSON-RPC sidecars in Rust and Go.
ClawBox
Open-Source Agent Framework
Plugin-based framework for building AI agents. Multi-model routing, tool orchestration, and execution loops with self-improvement hooks. Designed so each agent is a composition of swappable capabilities rather than a monolith.
VEIL Intelligence
Local AI Compute Stack
GPU-accelerated inference on local hardware — Ollama and vLLM serving models, ChromaDB for vector search, and RAG pipelines that stay entirely within the home lab. No cloud calls for AI workloads.
SlopeSniper
Algorithmic Trading System
Technical analysis engine with automated backtesting, real-time signal generation, and position sizing. Strict code quality — pytest, ruff, mypy enforced. Clear error messages when resource limits hit, with actionable fix instructions.
TakeOff Pro
Workflow Automation Platform
Production workflow orchestration handling CI/CD pipelines, deployment sequencing, and business process automation. Web dashboard for monitoring runs and configuring triggers.
Idea Park
Project Incubation SaaS
Platform for managing the pipeline from concept to shipped product. Stripe payment integration, user auth, structured stage gates. Built to track and prioritize across 40+ concurrent side projects.
What I Work With
Languages
Frontend
Backend
AI / ML
Blockchain
Infrastructure
DevOps
Platforms
Home Lab
Proxmox cluster — dedicated VMs for AI, agents, and production services
NVIDIA GPU serving local LLMs via Ollama and vLLM. No cloud inference dependency
Cloudflare Tunnels for ingress, NGINX reverse proxies, VLAN isolation between workloads
Get in touch.
Open to technical co-founder conversations, B2B projects involving AI or automation, and interesting infrastructure problems. Also happy to talk shop about agent architecture or home lab setups.
[email protected]© 2026 Mohsen Jahanshahi
Self-hosted on Docker · Cloudflare Tunnel · Next.js