Freelance AI · XR · Real-time systems
I'm Dimatri, a freelance engineer in Brussels. Companies bring me in to build LLM assistants, training simulators, and production platforms. Fixed scope, weekly demos, code you own.
Seven years turning generative AI into field-ready systems: from LLM-driven training avatars to police tactical simulators running in real deployments.
Defense & Police · Healthcare & First Aid · Enterprise & Safety
Built at and with: OneBonsai · Inoopa · Telemens
A sample of real-time AI and XR products I've architected and delivered. Each one had to stay responsive and reliable under live use.
AI · XR · OneBonsai · Role: Technical Director
Scripted NPCs made VR training predictable and flat. Trainees learned the script, not the skill.
Outcome: Built LLM-driven avatars that hold unscripted conversations, show emotion, and remember what a trainee did. Deployed across fire-safety, first-aid, and cybersecurity training.
Stack: Unreal Engine, C++, OpenAI API, NLP, Behavior Trees, TTS / STT
Simulation · AI · OneBonsai · Role: Lead Developer
Live tactical training is costly, risky, and impossible to repeat under identical conditions.
Outcome: Led a high-fidelity VR trainer with AI adversaries, room-clearing logic, and ballistics, where a trainee's decisions reshape the scenario. In use with law enforcement.
Stack: Unreal Engine, C++, Multi-Agent AI, State Machines, Ballistics Sim
Platform · XR · OneBonsai · Role: Founding Engineer
Every new VR scenario was a slow, bespoke build with no shared foundation to stand on.
Outcome: Founding engineer of the platform that let new modules ship fast on reusable frameworks, scaled across fire safety, first aid, cybersecurity, and paramedic training.
Stack: Unreal Engine, C++, Multi-User VR, Web APIs, Deployment Tooling
AI · Full-stack · Inoopa · Role: Technical Lead
Sales teams burn hours hunting for qualified companies across scattered, stale data.
Outcome: Owned an AI platform end to end: architecture, data pipelines that enrich company intelligence at scale, the full prospecting UI, and cloud infrastructure.
Stack: Next.js, TypeScript, FastAPI, Apache Airflow, Kubernetes, Terraform
AI · Web · Telemens · Role: Co-founder
AI image tools force a stop-start workflow: prompt, wait, break focus, repeat.
Outcome: Co-founded and built a real-time, multi-user canvas where AI generation lives inside the creative flow, from the design system to the LLM orchestration.
Stack: FastAPI, LangChain, OpenAI API, Vector DBs, Angular, WebSockets
Hardware · IoT · Open Source · Role: Founder
Assistive navigation tech is expensive and locked behind proprietary hardware.
Outcome: Designed an open-source haptic glove that scans obstacles up to 20 m away, with free build instructions, assistive tech anyone can make affordably.
Stack: C++, IoT, Product Design, Web
Timeboxed builds with a fixed scope, weekly demos, and a documented handoff. Pick the altitude that fits where you are.
Prove the risky idea, fast. One focused, working prototype that answers the question stopping you from committing.
You get: A working prototype build · A short demo video · Clear next-steps outline · A guided handoff session
Best for: New product ideas · High-risk features · VR / AR interaction tests · AI feature demos
Ship a v1 people can rely on. A production-ready first version, tested on the flows that matter, ready for real users.
You get: A production-ready v1 · Core user flows tested · Launch & setup guide · Support through your launch window
Best for: Web apps & internal tools · VR / AR applications · AI-powered workflows · Customer-facing products
A complete system, end to end. Architecture through deployment for a full system a team or enterprise runs on.
You get: A complete system build · Admin & operations tooling · Reporting or analytics layer · Deployment & onboarding pack
Best for: Enterprise platforms · Multi-scenario training · Advanced AI agents · Cross-team internal systems
No open-ended retainers or mystery timelines. You always know what's being built, when it ships, and what it costs.
We align on the goal, deadline, and constraints. Within 48 hours you get a short plan: what gets built, how success is measured, the timeline, and a fixed quote. You approve, we start.
Work runs in weekly cycles with a new build every Friday. You give feedback, it rolls into the next cycle, and you always have a live link to the latest version. No black boxes.
I ship into your environment, whether that's publishing the app, installing the VR build, or wiring up the automation, with a recorded walkthrough and clear docs. You leave able to run it without me.
I'm Dimatri Godunov, a full-stack engineer specializing in AI and XR systems. For seven years I've built the hard parts: real-time simulation, LLM-driven characters, and the backends that keep them stable, hands-on from first prototype to production.
I work solo by default, so you deal with the person building your product, not an account manager. When a project needs extra hands, I bring trusted collaborators and stay accountable for the result. My background is production software under pressure: systems shipped to defense, healthcare, and enterprise clients where "it mostly works" isn't an option.
Usually this week. We scope in 48 hours and the first build cycle starts right after you approve.
Yes. A fixed quote follows the 48-hour scope, so there are no surprises. I'm also open to hourly or monthly arrangements for ongoing work.
Unreal Engine, Unity, and Blender for 3D/VR/AR; C++ and C# for real-time systems; Python for AI pipelines; TypeScript with Next.js, React, and Angular for web. Plus LangChain and n8n for AI workflows and automation, deployed on major cloud platforms.
I do. You work directly with me from scope to ship. If a project needs more hands, I bring trusted collaborators and stay accountable for the outcome.
Either. I can ship into your environment and manage it, or hand over a documented build your team runs. We decide at scope.
Available as a light retainer or as follow-on build cycles, whatever the product needs once it's live.
Your goal and a rough deadline. We'll sort the technical details together during the scope call.
Book a 30-minute intro and we'll talk feasibility, approach, and timeline, no obligation. Prefer to write? Use the form and I'll reply within a day.
Brussels, Belgium · Working EN · NL · FR