3D Reconstruction & Generative AI E...

Bhatiyani Astute Intelligence is seeking a talented 3D Reconstruction & Generative AI Engineer to spearhead the development of an innovative end-to-end system. This role involves creating high-precision 3D jewellery models from both images and text prompts, pushing the boundaries beyond traditional computer vision. You will architect systems that seamlessly integrate:

• 3D Reconstruction: Utilizing multi-view and single-view techniques.
• Generative AI: Implementing text-to-3D and image-to-3D generation capabilities.
• RAG Pipelines: For efficient design retrieval and conditional generation.

The ultimate goal is to produce manufacturing-grade meshes suitable for advanced applications such as gold casting and 3D printing.

Key Responsibilities:

3D Reconstruction & Geometry:

• Develop robust pipelines for generating 3D meshes from multi-view and single-view inputs, leveraging learned priors.
• Implement advanced techniques for point cloud generation, fusion, meshing, and surface reconstruction/refinement.
• Ensure exceptional geometric accuracy (sub-millimeter level) and clean, watertight mesh topology.

Generative AI (Core Focus):

• Design, develop, and fine-tune state-of-the-art models for text-to-3D and image-to-3D generation.
• Work with cutting-edge methods including diffusion models (2D to 3D lifting), NeRF/Gaussian Splatting, and latent 3D representations.
• Integrate features for style control and conditional generation based on user constraints.

RAG (Retrieval-Augmented Generation) Systems:

• Build sophisticated RAG pipelines to retrieve existing designs and CAD assets.
• Condition generative models using these retrieved references for more accurate and relevant outputs.
• Design efficient embedding pipelines for images and text, and set up vector databases for rapid design search.
• Enable intuitive workflows such as 'Generate similar to this design'.
• Develop hybrid pipelines combining retrieval, conditioning, generation, and refinement.
• Optimize for low-latency retrieval and high semantic relevance.

Model Optimization & Deployment:

• Optimize generative models for GPU efficiency, including multi-GPU setups.
• Develop batch inference pipelines.
• Deploy generation APIs and scalable backend systems.

Post-Processing & CAD Readiness:

• Implement mesh cleanup routines, including hole filling and smoothing.
• Perform topology correction and conversion to CAD-friendly formats like STL/STEP.
• Ensure all outputs are print-ready and structurally sound for casting.

Required Skills:

• Core: Strong proficiency in Python and PyTorch. Deep understanding of 3D geometry (meshes, point clouds, SDFs), multi-view geometry, camera calibration, and differentiable rendering. Experience with PyTorch3D, Open3D, Trimesh, and various 3D reconstruction pipelines (SfM, MVS, NeRF).
• Generative AI: Hands-on experience with diffusion models (e.g., Stable Diffusion, latent diffusion), GANs, and 3D generative models. Understanding of text-to-image/text-to-3D pipelines, conditioning, and prompt engineering.
• RAG & LLM Systems: Proven experience building RAG pipelines. Familiarity with vector databases (e.g., FAISS, Milvus) and embedding models (e.g., CLIP, multimodal embeddings). Ability to integrate structured (CAD) and unstructured (images/text) data and design semantic search systems.

Preferred / Bonus Skills:

• Experience in jewellery design or CAD workflows. Proficiency with Blender, MeshLab, Rhino, or Fusion 360. Familiarity with vision-language models and grounding models. Experience with multi-GPU training and diffusion + 3D hybrid pipelines.

Expected Outcomes:

• A production-ready system capable of generating high-quality 3D jewellery from text or images.
• A RAG-powered design system for retrieval-assisted generation.
• Consistent, high-quality design outputs across various iterations.

Candidate Profile:

• 2-6+ years of experience in Computer Vision, Generative AI, or 3D Graphics.
• Strong ability to bridge research understanding with production deployment.

Nice-to-Have Project Experience:

• Text-to-3D systems, RAG-based generative pipelines, high-detail reconstruction (jewellery/mechanical parts), and CAD-integrated AI systems.

Location: Bengaluru, Karnataka, India.