Back to dashboard

Spectral is building generative foundation models that turn user input into stru…

completed37 qualified1 runApr 30, 4:55 AMspectral-is-building-generative-foundation-models-that-turn
ParsedSpectral · 6 topics · Senior · Hybrid
Generating seed nodes
0 proposed
Explored 0 queries
0/0 done
    3
    Expanding nodes
    queued
    4
    Qualifying candidates
    queued

    Qualified Candidates (23)

    BD

    Boyang Deng

    high hireability

    Student Researcher@DeepMind

    Previously: Research Scientist @ Waymo

    San Francisco, US

    50
    Geometric Deep Learning85
    Model Architecture Design78
    Training Methodology68
    Generative 3D CAD45
    CAD & Computational Geometry18
    RL for Generative Models5
    Strengths
    CvxNet (307 citations): learnable convex decomposition for 3D shape representation
    Shape programs with repeatable implicit parts — structured/parametric 3D geometry
    Gaps
    No explicit CAD/B-rep/NURBS/STEP or feature tree experience
    …click to see all
    BY

    Brent Yi

    high hireability

    PhD student@UC Berkeley

    Previously: Perception Intern @ Waymo

    San Francisco, US

    43
    Geometric Deep Learning75
    Training Methodology65
    Model Architecture Design62
    RL for Generative Models28
    Generative 3D CAD18
    CAD & Computational Geometry8
    Strengths
    Nerfstudio co-author (917-citation 3D neural field framework, SIGGRAPH 2023)
    gsplat co-author — production 3D Gaussian splatting library (157 citations)
    Gaps
    No structured parametric CAD generation — all 3D work is reconstruction, not synthesis
    …click to see all
    AK

    Aliasghar Khani

    medium hireability

    Researcher@Autodesk

    Previously: AI Research Internship @ Autodesk

    Vancouver, CA

    37
    Model Architecture Design58
    Training Methodology50
    Generative 3D CAD45
    Geometric Deep Learning45
    CAD & Computational Geometry20
    RL for Generative Models5
    Strengths
    WaLa: billion-param 3D wavelet-latent diffusion model (2024)
    Autodesk Research — 3D shape generation group, CAD-adjacent org
    Gaps
    WaLa uses implicit SDFs — not structured/parametric CAD (B-rep, NURBS, feature trees)
    …click to see all
    DR

    Daxuan Ren

    medium hireability

    Principal Software Development Engineer@Autodesk

    Previously: Algorithm Researcher @ SenseTime

    Singapore, SG

    71
    Generative 3D CAD92
    CAD & Computational Geometry90
    Geometric Deep Learning88
    Model Architecture Design80
    Training Methodology70
    RL for Generative Models5
    Strengths
    CVPR 2026: bidirectional query-driven parametric CAD sketch generation
    NeurIPS 2025: learning CAD modeling sequences — structured sequential output
    Gaps
    No RL for generative models evidence (RLHF/DPO/GRPO) — zero signal
    …click to see all
    ES

    Edward J. Smith

    medium hireability

    Machine Learning Researcher@Borealis AI

    Previously: Visiting Researcher @ Meta

    Toronto, CA

    49
    Geometric Deep Learning82
    Model Architecture Design72
    Training Methodology62
    Generative 3D CAD35
    CAD & Computational Geometry22
    RL for Generative Models20
    Strengths
    Kaolin (NVIDIA): core PyTorch 3D DL library, 118 citations
    GEOMetrics (ICML 2019): graph-encoded mesh generation architecture
    Gaps
    No CAD/B-rep/NURBS/parametric structured 3D experience
    …click to see all
    HY

    Hong-Xing Yu

    medium hireability

    PhD Candidate@Stanford University

    Previously: Intern @ Google

    US

    40
    Model Architecture Design82
    Training Methodology78
    Geometric Deep Learning62
    Generative 3D CAD10
    RL for Generative Models5
    CAD & Computational Geometry5
    Strengths
    WonderWorld (CVPR 2025): interactive 3D scene generation from single image
    PhysDreamer: physics-based 3D object interaction via video diffusion
    Gaps
    No CAD, B-rep, NURBS, or parametric geometry experience — core requirement
    …click to see all
    KM

    Kanika Madan

    medium hireability

    Principal Researcher@Autodesk

    Previously: Senior Manager Business Solutions @ AU

    Gurugram, IN

    52
    Model Architecture Design70
    RL for Generative Models65
    Training Methodology60
    Geometric Deep Learning55
    Generative 3D CAD40
    CAD & Computational Geometry20
    Strengths
    WaLa (2024): billion-param 3D diffusion model at Autodesk — core 3D gen work
    GFlowNets ICML 2023 (119 citations): RL-aligned generative model expertise
    Gaps
    3D work is SDF/implicit — no structured parametric CAD (B-rep, feature trees)
    …click to see all
    LS

    Liangchen Song

    medium hireability

    Engineer@Apple

    Previously: Intern @ InnoPeak Technology

    San Francisco, US

    36
    Model Architecture Design65
    Training Methodology60
    Geometric Deep Learning58
    Generative 3D CAD20
    RL for Generative Models5
    CAD & Computational Geometry5
    Strengths
    Cavia (ICLR 2025): multi-view video diffusion, View-Integrated Attention
    NeuRBF (ICCV 2023): adaptive radial basis neural fields for 3D
    Gaps
    No CAD, B-rep, NURBS, or parametric geometry experience
    …click to see all
    LL

    Lingxiao Li

    medium hireability

    Research Scientist@Netflix

    Previously: PhD student @ Massachusetts Institute of Technology

    41
    Geometric Deep Learning72
    Model Architecture Design60
    Training Methodology55
    Generative 3D CAD30
    CAD & Computational Geometry25
    RL for Generative Models5
    Strengths
    SPFN (CVPR 2019 oral, 279 cites) — geometric primitive fitting on 3D point clouds
    MIT CSAIL PhD, Justin Solomon lab — top computational geometry pedigree
    Gaps
    No B-rep/NURBS/parametric CAD or feature tree experience
    …click to see all
    MG

    Matheus Gadelha

    medium hireability

    Research Scientist@Adobe

    Previously: Research Assistant @ University of Massachusetts Amherst

    Seattle, US

    55
    Geometric Deep Learning92
    Model Architecture Design78
    Training Methodology65
    Generative 3D CAD55
    CAD & Computational Geometry30
    RL for Generative Models10
    Strengths
    GEM3D (SIGGRAPH 2024): generative 3D shape synthesis via medial abstractions
    Residual Primitive Fitting (CVPR 2026): structured shape → primitive decomposition
    Gaps
    No parametric CAD experience (B-rep, STEP, feature trees) — visual 3D, not structured editable CAD
    …click to see all
    PR

    Pradyumna Reddy

    medium hireability

    Senior Research Scientist@Autodesk

    Previously: Senior Research Scientist @ Huawei

    London, GB

    74
    Generative 3D CAD95
    CAD & Computational Geometry95
    Geometric Deep Learning92
    Model Architecture Design88
    Training Methodology70
    RL for Generative Models5
    Strengths
    DualBrep (SIGGRAPH 2026): dual-field continuous B-rep — exact CAD representation match
    WaLa: billion-parameter 3D generative model, wavelet latent diffusion at scale
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) evidence
    …click to see all
    AJ

    Alec Jacobson

    low hireability

    Researcher@Adobe

    Previously: Senior Research Scientist @ Adobe

    Toronto, CA

    53
    Geometric Deep Learning95
    Model Architecture Design62
    Training Methodology58
    CAD & Computational Geometry55
    Generative 3D CAD40
    RL for Generative Models5
    Strengths
    libigl creator — canonical geometry processing library, ~4k GitHub stars
    Neural Geometric Level of Detail (607 cites) — neural implicit 3D
    Gaps
    Tenured UofT professor — very unlikely to leave for startup
    …click to see all
    AR

    Arianna Rampini

    low hireability

    Senior Research Scientist@Autodesk

    Previously: Junior Research Scientist @ Autodesk

    IT

    60
    Geometric Deep Learning88
    Model Architecture Design82
    Training Methodology72
    Generative 3D CAD65
    CAD & Computational Geometry50
    RL for Generative Models5
    Strengths
    Make-A-Shape: 10M-scale 3D generative model (ICML 2024, 27 citations)
    WaLa: billion-param 3D diffusion, novel wavelet encodings (NeurIPS 2024)
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) in published work
    …click to see all
    BP

    Ben Poole

    low hireability

    Senior Staff Research Scientist@DeepMind

    Previously: Research Scientist @ Google

    San Francisco, US

    49
    Model Architecture Design93
    Training Methodology85
    Geometric Deep Learning62
    Generative 3D CAD38
    RL for Generative Models8
    CAD & Computational Geometry5
    Strengths
    DreamFusion: score distillation sampling for text-to-3D (3,287 citations)
    CAT3D (NeurIPS 2024 Oral) + CAT4D (CVPR 2025 Oral) — active 3D generative research
    Gaps
    All 3D work is NeRF/mesh/multi-view — zero parametric CAD, B-rep, or NURBS experience
    …click to see all
    DL

    Difan Liu

    low hireability

    Research Scientist@Adobe

    Previously: PhD student @ University of Massachusetts, Amherst

    65
    Geometric Deep Learning82
    Model Architecture Design78
    Generative 3D CAD72
    Training Methodology65
    CAD & Computational Geometry65
    RL for Generative Models28
    Strengths
    CSGNet (CVPR 2018, 279 citations): neural CSG parser for CAD primitives
    ParSeNet (ECCV 2020): parametric surface fitting — planes, spheres, NURBS on 3D data
    Gaps
    Recent work (2024-2025) shifted to video generation — 3D CAD not current focus
    …click to see all
    HS

    Hooman Shayani

    low hireability

    Principal Researcher@Autodesk

    London, GB

    75
    CAD & Computational Geometry97
    Geometric Deep Learning96
    Generative 3D CAD95
    Model Architecture Design82
    Training Methodology75
    RL for Generative Models5
    Strengths
    BRepNet (CVPR 2021, 159 citations) — topological GNN for B-rep solid models
    UV-Net (CVPR 2021, 125 citations) — learning from boundary representations
    Gaps
    No RL/RLHF/DPO publications — RL for generative models is a gap
    …click to see all
    KH

    Ka-Hei Hui

    low hireability

    Researcher@Autodesk

    Previously: PhD student @ The Chinese University of Hong Kong

    US

    50
    Geometric Deep Learning82
    Model Architecture Design78
    Training Methodology62
    Generative 3D CAD55
    CAD & Computational Geometry15
    RL for Generative Models5
    Strengths
    Make-A-Shape: 10M-scale 3D shape foundation model (2024, Autodesk AI Lab)
    Neural Wavelet Diffusion (144 cites): wavelet-domain 3D diffusion, novel rep
    Gaps
    No structured/parametric CAD (B-rep, feature trees, STEP) — free-form only
    …click to see all
    KK

    Karsten Kreis

    low hireability

    Principal Research Scientist@NVIDIA

    Previously: Senior Research Scientist II @ NVIDIA

    Vancouver, CA

    53
    Model Architecture Design92
    Training Methodology88
    Geometric Deep Learning75
    Generative 3D CAD55
    RL for Generative Models5
    CAD & Computational Geometry5
    Strengths
    Magic3D (1,461 cites) — text-to-3D, directly relevant 3D generative work
    LION: latent point diffusion for 3D shapes — NeurIPS 2022, 634 cites
    Gaps
    No CAD, B-rep, or parametric structure experience
    …click to see all
    MT

    Matthew Tancik

    low hireability

    Researcher@Luma AI

    Previously: PhD student @ University of California Berkeley

    42
    Training Methodology78
    Model Architecture Design78
    Geometric Deep Learning60
    Generative 3D CAD28
    RL for Generative Models5
    CAD & Computational Geometry5
    Strengths
    NeRF co-inventor (15k citations) — world-class 3D implicit representation expertise
    Nerfstudio: production NeRF training framework used industry-wide
    Gaps
    No CAD domain knowledge: B-rep, NURBS, feature trees, STEP files
    …click to see all
    NM

    Nigel J. W. Morris

    low hireability

    Autodesk

    Previously: Researcher @ Autodesk

    67
    Generative 3D CAD95
    CAD & Computational Geometry95
    Geometric Deep Learning85
    Model Architecture Design72
    Training Methodology50
    RL for Generative Models5
    Strengths
    SolidGen (TMLR 2023): autoregressive direct B-rep synthesis — core domain
    UV-Net (2020): B-rep neural architecture, operates on CAD topology
    Gaps
    No RL for generative models evidence (RLHF/DPO/GRPO)
    …click to see all
    RW

    Rundi Wu

    low hireability

    Research Scientist@DeepMind

    Previously: PHD Student @ Columbia University

    San Francisco, US

    70
    Generative 3D CAD97
    Model Architecture Design85
    Geometric Deep Learning82
    CAD & Computational Geometry78
    Training Methodology72
    RL for Generative Models5
    Strengths
    DeepCAD (ICCV 2021) — autoregressive generative model for CAD sequences
    Lead contributor to ChrisWu1997/DeepCAD (15 commits)
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) in any publications
    …click to see all
    TC

    Ta-Ying Cheng

    low hireability

    Research Scientist@Netflix

    Previously: Research Scientist Intern @ Stability AI

    San Francisco, US

    35
    Model Architecture Design65
    Geometric Deep Learning62
    Training Methodology55
    Generative 3D CAD22
    CAD & Computational Geometry5
    RL for Generative Models2
    Strengths
    3DMiner (ICCV 2023): 3D shape generation from unannotated images
    Continuous 3D Words (CVPR 2024): 3D-aware diffusion controls
    Gaps
    No CAD, B-rep, NURBS, or parametric structure experience
    …click to see all
    TG

    Thibault Groueix

    low hireability

    Research Scientist@Meta

    Previously: Research Scientist @ Adobe

    San Francisco, US

    53
    Geometric Deep Learning92
    Model Architecture Design88
    Training Methodology65
    Generative 3D CAD45
    CAD & Computational Geometry22
    RL for Generative Models5
    Strengths
    AtlasNet (CVPR 2018, 1678 citations) — seminal 3D surface generation architecture
    Meta AI RS: training large-scale 3D generative AI and world models
    Gaps
    No structured/parametric CAD background — no B-rep, NURBS, STEP, or feature-tree work
    …click to see all

    Runs

    #1completed0 qualified / 0 foundApr 30, 4:55 AM