We build AI and agent systems that ship to production — language-agnostic, not just demos.
Dev Art Solutions is a family-run engineering studio building production-oriented AI systems, developer tools, and automation platforms.
Our focus:
- LLM integrations into existing software
- Local / on-premise AI infrastructure (Ollama)
- Conversational and agentic systems
- Workflow and process automation
- Open-source developer tooling and SDKs
We care about practical AI that runs inside real products — across whatever stack the product needs (.NET, Node, React, Angular).
A desktop trading intelligence platform for MetaTrader 5 traders: R-multiple risk accounting, strategy-vs-execution compliance analysis, behavioral/emotional discipline detection (FOMO, revenge trading, overtrading), quantitative research (Monte Carlo, Risk of Ruin, bootstrap intervals), and interactive candlestick trade replay.
Stack: Python · PySide6 (Qt6) · SQLite / SQLAlchemy · MetaTrader5 SDK Repository: TradeAudit
A lightweight, strategy-agnostic MetaTrader 5 Expert Advisor that continuously monitors account risk and exposes deterministic SAFE, RESTRICTED, BLOCKED, and EMERGENCY states. It enforces configurable daily-loss, per-position risk, trade-count, open-position, trading-session, and spread policies with restart-safe state and opt-in scoped liquidation.
Stack: MQL5 · MetaTrader 5 · GitHub Actions Repository: MT5 RiskGuard
A local-first FastAPI service that receives structured trading signals, persists and de-duplicates them with a durable SQLite queue, applies a deterministic risk pipeline (stop-loss, spread, daily-loss, trade/position limits, risk-based position sizing), and routes approved orders to MetaTrader 5 with order_check()/order_send(), crash-safe reconciliation, and a full audit trail. Dry-run by default, with a dual opt-in required before any live order is sent.
Stack: Python · FastAPI · SQLite · MetaTrader5 SDK · GitHub Actions Repository: MT5 Execution Bridge
A platform for deliberate personal progress: long-term AI memory, task automation, and psychological profiling (IQ, EQ, Big Five) with real scoring logic. Integrates OpenClaw as an external agent service.
Stack: Next.js · PostgreSQL · Ollama / OpenAI · OpenClaw Live: samsaraforge.com
A learning platform that tracks where you make mistakes and automatically schedules practice on your weak points. Connects to OpenClaw to auto-generate content, code snippets, and diagrams.
Stack: Angular · PostgreSQL · AI quiz engine · OpenClaw Live: dev-smart.academy
An open-source, local-first, self-evolving multi-agent environment. Agents live inside a shared environment, communicate through an asynchronous message layer, acquire reusable skills, maintain persistent state, and can create new agent definitions. The v0.1 foundation isolates candidate generations and preserves last-known-good rollback metadata while promotion remains human-controlled.
EvoMesh works directly with Ollama and other local OpenAI-compatible endpoints; it can also use InferHub as an optional local/self-hosted model provider.
Repository: EvoMesh · Documentation
A .NET client library for integrating with Ollama APIs in C# applications — local LLM integrations, ASP.NET Core apps, streaming chat, and developer tooling.
Repository: OllamaClient
A self-hosted LLM inference mesh in .NET. One Ollama-compatible API in front, a pool of GPU worker nodes behind it — run the hub where you have no GPU, run the nodes where you do.
Useful for:
- pooling GPUs across several machines behind one stable endpoint
- serving LLMs from behind a home router, with no port forwarding
- letting existing Ollama clients, scripts, and editor plugins keep working unchanged
- pluggable inference backends (Ollama first; vLLM, llama.cpp and others can slot in later)
Built on our own OllamaClient.
Repository: InferHub
A set of typed .NET clients for talking to an InferHub coordinator from C#. Dependency-injection friendly, System.Text.Json under the hood, and designed for lightweight integration with InferHub services.
Repository: InferHub.Clients
An open-source conversational AI framework for building agents, intent recognition, rules, and conversation flows.
Repository: SmoothLingua · SmoothLingua.Api
We build AI systems that are reliable, extensible, maintainable, observable, and production-ready.
Tools we work with:
AI / Agents Local LLMs · Agent systems · RAG · Ollama · InferHub · OpenClaw
Backend .NET / ASP.NET Core · C# · Python · Node · REST APIs
Frontend React · Next.js · Angular · TypeScript
Infra Docker · CI/CD · PostgreSQL · SQLite