Part-Time AI Applied Scientist (MSc.+/PhD.)

SWAPP.AI - Part-Time AI Applied Scientist (MSc.+/PhD.)

Hybrid (Tel Aviv) · ~3 days/week (~60%)

Who We Are

• Revolutionizing Architecture with AI: We’re transforming how buildings are designed and documented - using advanced AI to automate and accelerate the entire architectural workflow.
• Smart Design Meets Automation: Our platform learns each firm’s design language, BIM standards, and workflows to generate precise, build-ready documentation in record time.
• Architects’ Creative Co-Pilot: We empower architects and designers to focus on creativity, while SWAPP handles the heavy lifting of production and coordination.
• Built for Real-World Impact: From concept to construction, SWAPP’s technology bridges the gap between design vision and practical execution.
• A Culture of Innovation and Collaboration: We’re a fast-growing team of architects, engineers, and AI experts redefining the future of building design together.

What you’ll do

• Own problems end-to-end: You will drive the full cycle from research and prototyping to production-ready, integrated code. This is not an academic role; your work ships.
• Build and innovate: Research, design, and deliver novel models in geometry processing, LLMs/NLP, graph neural networks, reinforcement learning, causal inference and multimodal reasoning.
• Integrate with core products: Your models will live inside our production repositories, and you’ll collaborate directly with engineers, designers, and architects to make them real.
• Leverage unique real-world data: Work with rich, proprietary, challenging, messy, complex, and extraordinarily multimodal datasets (text + drawings + 2D/3D + BIM + codes) that combine architectural complexity with real-world constraints.

Must-have skills

• Advanced Degree: MSc or PhD (in progress or completed) in CS/AI/Applied-Math or a related field.
• Strong Programmer: Excellent Python skills, with experience writing clean, testable, and reproducible code.
• Proven Algorithm & ML Expertise: Deep fluency in algorithm design, model architectures & training, novel data representations, and data modeling (not just library use).
• Tech Stack Mastery: Hands-on with PyTorch and at least one of: GNNs, RL, causal methods, vision/geometry (OpenCV/Shapely), LLMs/NLP.
• Production Savvy: Experience deploying, monitoring, and maintaining ML models. 

Nice-to-Haves

• Familiarity with BIM/CAD or other structured, multimodal datasets.
• Intuition for 2D/3D geometry and spatial reasoning.
• Publications or impactful contributions (e.g., NeurIPS/ICLR/ICML/SoCG or open-source/industry projects).

Who thrives here

Our algorithms span many disciplines so we realize you may not be familiar with all of them. 

Therefore, you should be:

• Independent Team Player: You can learn domain knowledge by asking the right questions and execute on your own;
• Clear Communicator: You explain ideas, decisions, and code effectively to diverse collaborators.
• Builder-Researcher Mindset: relentlessly curious and ownership-driven, rigorous with software best practices-equally at home cracking novel algorithms and shipping clean, reliable production code.

What You Won’t Be Doing

• Tweaking hyper-parameters on standard datasets (this is custom problem domains)
• Chasing ML accuracy vanity metrics (geometric validity and architectural correctness matter more)
• Implementing papers blindly (you’ll be adapting and inventing)

Logistics & Growth

• Hybrid (Tel Aviv HQ) is the default; special circumstances may be considered.
• Commitment: ~3 days/week (~60%). We expect close collaboration and code review on your working days to keep you integrated with the team.
• Team & Mentorship: You'll be part of a senior, multidisciplinary team, working directly with (and mentored by) our core AI researchers, engineers, and product leaders.
• Growth: This is a core role with a clear path to a full-time position post-graduation for high-performers.

Apply: Send a your CV along with 1-2 examples describing hard problems you’ve tackled and the results or impact to [email protected]