L2labs
Turn ideas into AI products.
L2labs is a custom AI development studio from Europe focused on computer vision, 3D, LLMs & generative AI.
We help companies go from vague AI ideas to production-ready products – not research slides.
🌍 Website: https://l2labs.dev
📫 Email: alexpatuk@l2labs.dev
🧠 Areas: Computer Vision · 3D / Spatial · LLMs & RAG · Document AI · MLOps
What we do
We design, build and ship AI systems that run in the real world:
Computer Vision & 3D
- PPE and safety monitoring
- Deepfake / media authenticity checks
- Medical image processing
- BIM / Matterport pipelines, 3D segmentation & digital twins
- Gaussian Splatting for immersive scenes and virtual tours
LLMs & Agentic AI
- RAG chatbots over private data
- Domain-specific assistants and tools
- Workflow automation and evaluation loops
Document AI
- OCR pipelines and PDF understanding
- Semantic search and Q&A over large document sets
- Redaction / anonymization and compliance-oriented workflows
Production MLOps
- Cloud, edge and mobile deployments
- Monitoring, retraining loops, cost & latency optimization
- CI/CD for models and data pipelines
Selected products & case studies
Documents AI – cross-platform desktop app (macOS / Windows / Linux) for
local, privacy-first document analysis:
- Vector-based semantic search over PDFs, DOCX, TXT, Markdown
- Chat with your documents using GPT-powered assistants
- Local processing by default with optional cloud features
- Project-based organization and high-performance indexing
🔗 https://documents.l2labs.dev
Construction Drones & 3D Analytics
- Autonomous drone data capture for construction sites
- 2D/3D reconstruction, BIM alignment and as-built checks
- AI-driven progress tracking, safety monitoring and quality insights
🔗 https://360.l2labs.dev
3D / Gaussian Splatting & Matterport Pipelines
- Turning 360° imagery, LiDAR, Matterport captures and video into
interactive 3D scenes
- Experiments and internal tools around Gaussian Splatting for
photorealistic virtual tours, digital twins and AR/VR experiences
How we work
We’re optimised for shipping, not just experimenting:
Clarify the problem
We work with your team to define what “success” actually means in metrics, UX and business impact.
Validate feasibility early
Rapid prototyping to confirm that the idea can be delivered within your constraints (data, budget, timeline).
Build production-ready systems
Clean APIs, clear contracts, monitoring, and deployment pipelines from the start.
Iterate with real users
We treat models as living systems – integrating feedback, retraining on new data and continuously improving.
This organization on Hugging Face
On this Hugging Face org you can expect:
- Open-sourced building blocks we use in projects (vision models, 3D utilities, document AI tools, evaluation scripts).
- Demo Spaces showcasing:
- Document search & chat over PDFs
- 3D / Gaussian Splatting visualizations and tooling
- Computer vision pipelines (detection, segmentation, tracking).
- Datasets (where licensing allows) for benchmarking and experimentation.
If you’re building something in:
- 3D / spatial computing
- Document intelligence & anonymization
- Vision-heavy products (PPE, retail, OOH ads, safety, industrial)
- LLM copilots on top of complex internal data
— we’d be happy to chat.
Work with us
If you’re interested in collaborating on open models or hosting your own
demo here under the L2labs umbrella, reach out and tell us what you’re building.