3D Artist - OpenUSD & Synthetic Data · AI/ Computer Vision
⚲ Wrocław
12 000 - 15 000 PLN netto (B2B)
Wymagania
- 3D modeling
- OpenUSD
- Blender
- Synthetic data generation
- Python
- Texturing
Opis stanowiska
About Surveily
Surveily is the AI safety and operations intelligence company for industrial workplaces. We build our own AI and video-analytics product that turns the cameras companies already have into an intelligent system - detecting hazards in real time and preventing incidents before they happen. We operate at the intersection ofAI and Industry 4.0, solving a high-priority business problem around workplace safety and operational efficiency. Our platform integrates with 95% of existing digital camera infrastructure, and our customers have seen a 1152% increase in unsafe-behaviour visibility, 72% fewer safety incidents, and a 25% reduction in LTIR.
Now we are scaling, and the 3D content you create will directly train the computer-vision models that keep real people safe on real industrial sites.
About the role
Our object-detection models (YOLO) are trained in large part on synthetic data: photorealistic 3D scenes of industrial environments - warehouses, production halls, yards - populated with people, forklifts, machinery, and PPE. The pipeline already exists and runs end-to-end: a Python-native SimulationApp drives Synthetic Data Generation from an established OpenUSD scene and asset library. You are not starting from zero. What we need is the person who owns this pipeline and makes it dependable: creating new assets in Blender, maintaining and extending the existing USD library, and running SDG batches that deliver clean, training-ready datasets to the ML team - on time, every time.
The challenge
Your work has downstream consumers - automated pipelines and ML engineers - so a broken hierarchy, a missing semantic label, or a scene that "works on my machine" has a real cost: failed generation runs, wasted GPU hours, and models that miss hazards in the field. This is not a portfolio-art role. Visual quality matters, but consistency, technical correctness, and predictable delivery matter more. We are looking for someone who treats the asset library like production code - versioned, validated, documented - and treats a dataset handed off on schedule as the win.
What you'll own
Asset Creation (Blender)
• Model, texture, and optimise new 3D assets - industrial equipment, vehicles (forklifts, trucks), people, PPE, props, and environment elements.
• Export to USD cleanly and consistently: correct scale, pivots, hierarchies, naming, materials, and the semantic labels SDG annotation depends on.
• Keep polycount, texture budgets, and LODs fit for large-scale batch rendering.
USD Library Stewardship
• Edit, repair, and extend existing OpenUSD assets and scenes - composition arcs (references, payloads, variants), layering, materials, semantics.
• Maintain a versioned, documented library with predictable structure, so anyone on the team can reuse assets without surprises.
• Validate every asset before it enters the library - catch broken references, wrong units, or missing semantics before a failed training run does.
Synthetic Data Generation
• Own and operate the SDG process end to end: configure and launch generation batches - scene selection, domain randomisation (lighting, camera poses, textures, placement), annotation outputs for YOLO training.
• Monitor runs, verify output and label quality, and hand off clean, documented datasets to the ML team on schedule.
• Close the loop with ML engineers: when a model underperforms on a class, translate that into concrete asset or randomisation changes and re-generate.
What you'll bring
• 3+ years of professional 3D experience (games, VFX, simulation, archviz, or synthetic data) with strong Blender skills - modelling, UVs, texturing, materials.
• Hands-on OpenUSD experience. You understand composition (references, variants, layers), can debug a broken stage, and know why a mangled hierarchy or wrong up-axis breaks a pipeline.
• Working Python. Enough to run, configure, and lightly modify pipeline scripts - you'll be driving the SDG app, not building it.
• Computer-vision awareness. You understand what makes synthetic data useful for detection models: annotation types, domain randomisation, the sim-to-real gap.
• High ownership and discipline. You plan your work, communicate status honestly, flag problems early, and finish what you start - a track record of reliable, selfdirected delivery.
• Communicative English and Polish for day-to-day collaboration with the team.
Nice to have
• Experience with NVIDIA Omniverse, Isaac Sim, or Replicator specifically. Experience training or evaluating detection models (YOLO or similar) - even at hobby level.
• PBR workflows, Substance Painter/Designer, or photogrammetry.
• Git or other version control for asset management.
• Comfort working with AI assistants (Claude, ChatGPT, Copilot) in your daily workflow - for scripting, debugging, and learning new tools fast.
• Exposure to industrial, logistics, or safety environments.
What we offer
• A mission with real impact- the datasets you generate train models that prevent injuries on real industrial sites.
• Full ownership of the 3D content and SDG pipeline - your improvements ship straight into production models.
• A proven, working pipeline - you are not starting from zero.
• Direct collaboration with the ML team and founders; short feedback loops, no bureaucracy. An LLM-backed house - you get full access to frontier AI models (Claude and others) as a daily co-pilot for Python scripting, USD debugging, and Omniverse work, so you're never stuck alone on the technical side.
• Hardware capable of the job (workstation-class GPU) and the tools you need.
• Hybrid work from our Wrocław office.
• A B2B contract with compensation matched to experience.
Recruitment process
We will contact only selected candidates whose CV and portfolio clearly demonstrate the skills we are looking for. The process consists of a CV and portfolio review, a 5-minute phone call, a 30-minute video interview, and a practical task — modelling a small asset in Blender, exporting it to USD to our spec, and walking us through your validation checks. We will get back to you at the end of each stage.
Please include with your application: a portfolio or asset samples, and - if possible - an example of USD or pipeline work.
If you have more questions, please email us at: talent@surveily.com
Surveily is the AI safety and operations intelligence company for industrial workplaces. We build our own AI and video-analytics product that turns the cameras companies already have into an intelligent system - detecting hazards in real time and preventing incidents before they happen. We operate at the intersection ofAI and Industry 4.0, solving a high-priority business problem around workplace safety and operational efficiency. Our platform integrates with 95% of existing digital camera infrastructure, and our customers have seen a 1152% increase in unsafe-behaviour visibility, 72% fewer safety incidents, and a 25% reduction in LTIR.
Now we are scaling, and the 3D content you create will directly train the computer-vision models that keep real people safe on real industrial sites.
About the role
Our object-detection models (YOLO) are trained in large part on synthetic data: photorealistic 3D scenes of industrial environments - warehouses, production halls, yards - populated with people, forklifts, machinery, and PPE. The pipeline already exists and runs end-to-end: a Python-native SimulationApp drives Synthetic Data Generation from an established OpenUSD scene and asset library. You are not starting from zero. What we need is the person who owns this pipeline and makes it dependable: creating new assets in Blender, maintaining and extending the existing USD library, and running SDG batches that deliver clean, training-ready datasets to the ML team - on time, every time.
The challenge
Your work has downstream consumers - automated pipelines and ML engineers - so a broken hierarchy, a missing semantic label, or a scene that "works on my machine" has a real cost: failed generation runs, wasted GPU hours, and models that miss hazards in the field. This is not a portfolio-art role. Visual quality matters, but consistency, technical correctness, and predictable delivery matter more. We are looking for someone who treats the asset library like production code - versioned, validated, documented - and treats a dataset handed off on schedule as the win.
What you'll own
Asset Creation (Blender)
• Model, texture, and optimise new 3D assets - industrial equipment, vehicles (forklifts, trucks), people, PPE, props, and environment elements.
• Export to USD cleanly and consistently: correct scale, pivots, hierarchies, naming, materials, and the semantic labels SDG annotation depends on.
• Keep polycount, texture budgets, and LODs fit for large-scale batch rendering.
USD Library Stewardship
• Edit, repair, and extend existing OpenUSD assets and scenes - composition arcs (references, payloads, variants), layering, materials, semantics.
• Maintain a versioned, documented library with predictable structure, so anyone on the team can reuse assets without surprises.
• Validate every asset before it enters the library - catch broken references, wrong units, or missing semantics before a failed training run does.
Synthetic Data Generation
• Own and operate the SDG process end to end: configure and launch generation batches - scene selection, domain randomisation (lighting, camera poses, textures, placement), annotation outputs for YOLO training.
• Monitor runs, verify output and label quality, and hand off clean, documented datasets to the ML team on schedule.
• Close the loop with ML engineers: when a model underperforms on a class, translate that into concrete asset or randomisation changes and re-generate.
What you'll bring
• 3+ years of professional 3D experience (games, VFX, simulation, archviz, or synthetic data) with strong Blender skills - modelling, UVs, texturing, materials.
• Hands-on OpenUSD experience. You understand composition (references, variants, layers), can debug a broken stage, and know why a mangled hierarchy or wrong up-axis breaks a pipeline.
• Working Python. Enough to run, configure, and lightly modify pipeline scripts - you'll be driving the SDG app, not building it.
• Computer-vision awareness. You understand what makes synthetic data useful for detection models: annotation types, domain randomisation, the sim-to-real gap.
• High ownership and discipline. You plan your work, communicate status honestly, flag problems early, and finish what you start - a track record of reliable, selfdirected delivery.
• Communicative English and Polish for day-to-day collaboration with the team.
Nice to have
• Experience with NVIDIA Omniverse, Isaac Sim, or Replicator specifically. Experience training or evaluating detection models (YOLO or similar) - even at hobby level.
• PBR workflows, Substance Painter/Designer, or photogrammetry.
• Git or other version control for asset management.
• Comfort working with AI assistants (Claude, ChatGPT, Copilot) in your daily workflow - for scripting, debugging, and learning new tools fast.
• Exposure to industrial, logistics, or safety environments.
What we offer
• A mission with real impact- the datasets you generate train models that prevent injuries on real industrial sites.
• Full ownership of the 3D content and SDG pipeline - your improvements ship straight into production models.
• A proven, working pipeline - you are not starting from zero.
• Direct collaboration with the ML team and founders; short feedback loops, no bureaucracy. An LLM-backed house - you get full access to frontier AI models (Claude and others) as a daily co-pilot for Python scripting, USD debugging, and Omniverse work, so you're never stuck alone on the technical side.
• Hardware capable of the job (workstation-class GPU) and the tools you need.
• Hybrid work from our Wrocław office.
• A B2B contract with compensation matched to experience.
Recruitment process
We will contact only selected candidates whose CV and portfolio clearly demonstrate the skills we are looking for. The process consists of a CV and portfolio review, a 5-minute phone call, a 30-minute video interview, and a practical task — modelling a small asset in Blender, exporting it to USD to our spec, and walking us through your validation checks. We will get back to you at the end of each stage.
Please include with your application: a portfolio or asset samples, and - if possible - an example of USD or pipeline work.
If you have more questions, please email us at: talent@surveily.com
🔍 Dekoder Ogłoszenia
🟡
You are not starting from zero.
Istnieje już pewna infrastruktura i procesy, ale niekoniecznie są one w pełni rozwinięte lub zoptymalizowane.
🔴
What we need is the person who owns this pipeline and makes it dependable
Oczekuje się, że będziesz odpowiedzialny za utrzymanie i poprawę istniejącego procesu, co może oznaczać rozwiązywanie problemów i wprowadzanie znaczących zmian.
🔴
creating new assets in Blender, maintaining and extending the existing USD library, and running SDG batches that deliver clean, training-ready datasets to the ML team - on time
Zakres obowiązków jest szeroki i obejmuje zarówno tworzenie, jak i zarządzanie, a także zapewnienie terminowości dostaw, co może być wymagające.