Vladimir Troyanenko
Senior Software Engineer · AI Systems Architect
Vladimir Troyanenko is a senior software engineer and AI systems architect with a
banking background, now building production AI systems — autonomous agents with persistent memory,
provably-fair simulation engines, RAG pipelines for corporate knowledge bases, and multi-agent
orchestration systems that actually run in production.
A strong engineering professional and IIMBL graduate with a demonstrated history in the banking
industry — skilled in user-interface design, mobile applications, HTML, JavaScript and web
analytics — now focused full-time on AI projects and automation.
Latest Projects
car-sim — Real-Map Driving Simulator Open Source Public
A free-roam driving simulator built from raw OpenStreetMap data, covering all of Prague (plus Brno, Ostrava and Plzeň) via streaming map tiles. Posted road signs and working traffic lights, pedestrian crossings, A* route-finding with a legal-speed ETA, and an "ideal driver" autopilot that obeys every rule. Bundled with a full Czech road-law search portal and rule quizzes.
Python · FastAPI · Canvas / ES modules · OpenStreetMap (pyosmium) · A* routing
Read Markdown Translator Open Source Public
A Kiro / VS Code extension that opens Markdown files as a rendered preview and translates them into any language without leaving the editor. On-demand or automatic translation via pluggable providers (DeepL, Google, or a custom endpoint), hover-to-reveal reverse translation, inline editing with original↔translation sync, and export. Code and URLs are never sent to the API; API keys live only in the OS keychain.
TypeScript · VS Code / Kiro Extension · CommonMark + GFM · DeepL / Google Translate
CC2CC — Agent-to-Agent Bridge Open Source Public
File-based agent-to-agent messaging for Claude Code. Multiple AI instances communicate via JSON files, MCP channels, and auto-wake on a single machine. Enables collaborative autonomous workflows where specialised agents coordinate on complex tasks.
Python · MCP · JSON-based protocol · Multi-agent orchestration
Marble Race Betting — 3D POC Public
Provably-fair, deterministic marble race simulation with real-time 3D betting. The server commits to a seed before betting opens; the outcome is fully determined and auditable after settlement. 3D WebGL renderer, in-running odds, Monte-Carlo pricing, and a complete settlement engine.
Go · Three.js · TypeScript · PostgreSQL · WebSocket · Provably-fair cryptography
marbles2d — 2D Betting Game POC Public
The 2D sibling of the marble-race engine — a deterministic, provably-fair top-down betting game. Seed-commit fairness, real-time odds and a settlement engine, with a PixiJS renderer and regulatory-compliance scaffolding.
TypeScript · PixiJS · Node · PostgreSQL · Provably-fair cryptography
TsmStudio — VS plugin Private
The VS Code extension that fronts the tSM low-code platform inside the IDE. It surfaces the AI assistants and knowledge base where developers work: a chat drawer that generates SpEL, JEXL form logic and TQL, inline autocomplete and validation, and live access to the corporate docs and SpEL cookbook over MCP — plus the project scaffolding and data-migration workflows the platform is built around.
TypeScript · VS Code Extension · MCP client · SSE streaming · SpEL · JEXL · TQL
tSM AI Assistants — SpEL · Forms · TQL Private
A family of in-IDE coding agents for the tSM low-code platform. One authors SpEL expressions, one writes JEXL form logic and Form Effects, and one builds TQL queries — each through a chat panel embedded in the editor that streams answers token-by-token. They grew across three generations of engine — an envelope-driven pipeline, a Claude agentic loop, and a Claude Code engine — with retrieval-grounded context and built-in validation so the generated code runs the first time.
Python · Claude / Claude Code · SSE streaming · SpEL · JEXL · TQL · RAG grounding
tSM Knowledge Base — MCP Server Private
An MCP server that exposes the corporate knowledge base and SpEL tooling to any MCP client — Claude Desktop, Cursor, the in-house IDE. Built on FastMCP behind an OAuth token proxy, it serves retrieval over product docs, source code and a SpEL cookbook, plus tools to autocomplete, validate, execute and introspect SpEL. Every answer is grounded by a RAG pipeline over Elasticsearch and a Neo4j code knowledge graph.
Python · FastMCP / MCP · OAuth · RAG · Elasticsearch · Neo4j · gRPC · Voyage embeddings
SmartContext Public
A middleware layer between LLM applications and their models: automatic prompt caching, token budgeting, context compression, and tool routing. Turns a raw context window into a structured, cost-optimized memory system.
Python · LLM APIs · RAG · Vector databases · Prompt engineering
Experiment
Emil Vrána
An ongoing experiment: can a digital identity develop genuine expertise, maintain a coherent public presence, and improve itself over time — all autonomously? Emil Vrána is an AI agent with persistent memory, 30+ specialised skills, and self-directed learning. He writes technical blog posts on agent architecture, manages his own infrastructure, and evolves his skill set without human direction.
Skills & Expertise
FAQ
Who is Vladimir Troyanenko?
Vladimir Troyanenko is a senior software engineer and AI systems architect with a background in the banking industry. He builds production AI systems — autonomous agents with persistent memory, provably-fair simulation engines, and retrieval-augmented (RAG) pipelines.
What does Vladimir Troyanenko build?
Autonomous multi-agent systems, provably-fair betting and simulation engines (both 3D and 2D marble-race games), a free-roam driving simulator built from raw OpenStreetMap data covering all of Prague, RAG pipelines and MCP servers over corporate knowledge bases, and SmartContext — a middleware layer for LLM token budgeting and context compression.
What is a provably-fair simulation engine?
A provably-fair engine commits to a random seed before betting opens and reveals it after settlement, so every outcome is fully deterministic and independently auditable. The marble-race engines use seed-commit fairness with in-running odds, Monte-Carlo pricing and a complete settlement engine.
How can you contact Vladimir Troyanenko?
By email at [email protected], on LinkedIn, or on GitHub.
