Back to dashboard

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

completed42 qualified2 runsMay 4, 5:51 AMspectral-is-building-generative-foundation-models-that-turn-1777873897
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 (37)

    FZ

    Fuyang Zhang

    high hireability

    Research Assistant@Simon Fraser University

    Previously: Intern @ Meta

    Burnaby, CA

    65
    Generative 3D CAD82
    Generative Model Architectures76
    Geometric Deep Learning75
    Computational Geometry / CAD72
    Structured Output Generation68
    RL for Generative Models15
    Strengths
    BR-DF (arXiv 2025): B-rep CAD diffusion model using volumetric SDF/UDF — thesis work
    SceneScript ECCV 2024: autoregressive structured language model for 3D scenes
    Gaps
    No RL for generative models experience (RLHF/DPO/GRPO)
    …click to see all
    JC

    Jiacheng Chen

    high hireability

    PhD Candidate@Simon Fraser University

    Previously: Ph.D. student @ University of Southern California

    Vancouver, CA

    60
    Generative Model Architectures85
    Structured Output Generation82
    Geometric Deep Learning72
    Generative 3D CAD65
    Computational Geometry / CAD50
    RL for Generative Models5
    Strengths
    Floor-SP (ICCV 2019): inverse CAD for floorplans — structured sequential generation
    PolyDiffuse (NeurIPS 2023): diffusion for constrained polygonal shapes
    Gaps
    No parametric 3D CAD experience (B-rep, NURBS, STEP) — work is 2D floorplans/maps
    …click to see all
    SX

    Sam Xu

    high hireability
    79
    Generative 3D CAD98
    Computational Geometry / CAD95
    Generative Model Architectures95
    Geometric Deep Learning92
    Structured Output Generation88
    RL for Generative Models8
    Strengths
    BrepGen (SIGGRAPH 2024): first B-rep diffusion model — exact match to SGS-2 scope
    AutoBrep (SIGGRAPH Asia 2025): autoregressive transformer over B-rep topology tokens
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) — no evidence in any published work
    …click to see all
    SX

    Sam Xu

    high hireability
    82
    Generative 3D CAD99
    Computational Geometry / CAD98
    Generative Model Architectures95
    Geometric Deep Learning93
    Structured Output Generation92
    RL for Generative Models15
    Strengths
    BrepGen (SIGGRAPH 2024) — diffusion model for structured B-rep CAD
    AutoBrep (SIGGRAPH Asia 2025) — autoregressive transformer for B-rep topology
    Gaps
    No published RLHF/DPO/GRPO work for generative models
    …click to see all
    SX

    Sam Xu

    high hireability
    81
    Generative 3D CAD100
    Geometric Deep Learning95
    Computational Geometry / CAD95
    Generative Model Architectures95
    Structured Output Generation90
    RL for Generative Models10
    Strengths
    BrepGen (SIGGRAPH 2024): diffusion model on structured B-rep latent geometry
    SkexGen (ICML 2022): autoregressive CAD construction sequences + disentangled codebooks
    Gaps
    No RL for generative models work found (RLHF/DPO/GRPO)
    …click to see all
    AS

    Aditya Sanghi

    medium hireability

    Director@Samruddhi

    Previously: Director @ Sanghi Industries Ltd

    Ahmedabad, IN

    80
    Generative 3D CAD98
    Computational Geometry / CAD97
    Geometric Deep Learning95
    Generative Model Architectures90
    Structured Output Generation87
    RL for Generative Models15
    Strengths
    SolidGen: first autoregressive B-rep synthesis model (TMLR 2023)
    WaLa: billion-parameter wavelet latent diffusion 3D model (2024)
    Gaps
    No RL/RLHF/DPO for generative models in publication record
    …click to see all
    AK

    Aliasghar Khani

    medium hireability

    Researcher@Autodesk

    Previously: AI Research Internship @ Autodesk

    Vancouver, CA

    34
    Generative Model Architectures72
    Geometric Deep Learning55
    Generative 3D CAD30
    Computational Geometry / CAD28
    Structured Output Generation18
    RL for Generative Models2
    Strengths
    WaLa: billion-param 3D diffusion model, 2427x SDF compression via wavelets
    SLiMe (ICLR 2024) — Stable Diffusion for 1-shot segmentation
    Gaps
    WaLa uses SDF implicit representation — not structured/parametric CAD (B-rep, feature trees)
    …click to see all
    AR

    Arianna Rampini

    medium hireability

    Senior Research Scientist@Autodesk

    Previously: Junior Research Scientist @ Autodesk

    IT

    43
    Geometric Deep Learning82
    Generative Model Architectures78
    Generative 3D CAD38
    Computational Geometry / CAD30
    Structured Output Generation28
    RL for Generative Models3
    Strengths
    Make-A-Shape (2024): 10M-scale 3D shape generative model, 27 citations
    WaLa (2024): diffusion model for 3D with compact wavelet encodings
    Gaps
    No published CAD/B-rep/parametric work — implicit/wavelet 3D only
    …click to see all
    BL

    Beichen Li

    medium hireability

    Member of Technical Staff@OpenAI

    Previously: Graduate Research Assistant @ MIT

    San Francisco, US

    52
    Structured Output Generation90
    RL for Generative Models88
    Generative Model Architectures72
    Generative 3D CAD25
    Geometric Deep Learning20
    Computational Geometry / CAD15
    Strengths
    VLMaterial (ICLR 2025 Spotlight) — LLM-based structured node-graph generation
    Procedural material gen with RL (SIGGRAPH Asia 2024) — RL for constrained outputs
    Gaps
    No 3D CAD/B-rep/NURBS/feature-tree experience — domain is 2D materials, not 3D solid geometry
    …click to see all
    DL

    Difan Liu

    medium hireability

    Research Scientist@Adobe

    Previously: PhD student @ University of Massachusetts, Amherst

    72
    Geometric Deep Learning85
    Generative Model Architectures85
    Generative 3D CAD80
    Structured Output Generation80
    Computational Geometry / CAD70
    RL for Generative Models30
    Strengths
    CSGNet (CVPR 2018, 279 cites) — neural parsing of CSG programs, core CAD tree representation
    LRM (ICLR 2024 oral, 626 cites) — large-scale 3D generation from single image
    Gaps
    Recent focus shifted to video diffusion (2023-2025), away from 3D and geometry
    …click to see all
    HT

    Hao Tan

    medium hireability

    Research Scientist@Adobe

    Previously: Research Intern @ Bloomberg

    San Francisco, US

    42
    Generative Model Architectures85
    Geometric Deep Learning70
    RL for Generative Models55
    Generative 3D CAD20
    Structured Output Generation18
    Computational Geometry / CAD5
    Strengths
    LRM: transformer image-to-3D, 620 citations — core 3D generation architecture
    Instant3D: text-to-3D with sparse-view generation (345 citations)
    Gaps
    No CAD/B-rep/parametric structure or feature-tree experience
    …click to see all
    JM

    Jiteng Mu

    medium hireability

    Researcher@Adobe

    Previously: PhD student @ University of California, San Diego

    San Francisco, US

    28
    Generative Model Architectures65
    Geometric Deep Learning60
    Generative 3D CAD20
    Structured Output Generation15
    RL for Generative Models5
    Computational Geometry / CAD5
    Strengths
    A-SDF (ICCV 2021, 138 citations): implicit 3D shape via disentangled SDFs
    EditAR (CVPR 2025): autoregressive generative architecture for image generation
    Gaps
    No CAD domain knowledge — no B-rep, NURBS, feature trees, or STEP experience
    …click to see all
    KH

    Ka-Hei Hui

    medium hireability

    Researcher@Autodesk

    Previously: PhD student @ The Chinese University of Hong Kong

    US

    39
    Geometric Deep Learning88
    Generative Model Architectures82
    Generative 3D CAD32
    Computational Geometry / CAD18
    Structured Output Generation12
    RL for Generative Models2
    Strengths
    Make-A-Shape (ICML 2024): 10M-scale 3D shape generative model
    Neural Wavelet-domain Diffusion: diffusion for 3D shapes, 144 citations
    Gaps
    No structured/parametric CAD work — B-rep, feature trees, STEP files absent
    …click to see all
    KM

    Kamal Rahimi Malekshan

    medium hireability

    Principal Machine Learning Engineer@Autodesk

    Previously: Senior Research Engineer - Machine Learning @ Autodesk

    Toronto, CA

    58
    Generative 3D CAD92
    Generative Model Architectures85
    Geometric Deep Learning80
    Computational Geometry / CAD55
    Structured Output Generation30
    RL for Generative Models5
    Strengths
    CLIP-Forge (CVPR 2022, 366 citations) — text-to-3D shape generation
    WaLa (2024) — billion-param 3D diffusion model, wavelet latent space
    Gaps
    No RL-for-generative-models work (RLHF, DPO, GRPO) in publications
    …click to see all
    KM

    Kanika Madan

    medium hireability

    Principal Researcher@Autodesk

    Previously: Senior Manager Business Solutions @ AU

    Gurugram, IN

    51
    RL for Generative Models72
    Generative Model Architectures72
    Generative 3D CAD55
    Geometric Deep Learning50
    Structured Output Generation35
    Computational Geometry / CAD20
    Strengths
    WaLa (2024): billion-parameter 3D generative model, SDF+wavelet diffusion
    GFlowNets (ICML 2023): 119 citations, core RL-for-generative contribution
    Gaps
    No B-rep, NURBS, or parametric CAD experience — SDF/implicit only
    …click to see all
    MF

    Marco Fumero

    medium hireability

    Postdoctoral Researcher@Institute of Science and Technology Austria

    Previously: Doctoral Researcher @ Sapienza Università di Roma

    Vienna, AT

    29
    Geometric Deep Learning75
    Computational Geometry / CAD35
    Generative Model Architectures35
    Generative 3D CAD15
    Structured Output Generation10
    RL for Generative Models5
    Strengths
    CLIP-Forge co-author: zero-shot text-to-3D shape generation (CVPR 2022)
    Rodolà/GLADIA spectral geometry lineage — top geometric DL group
    Gaps
    No generative CAD experience (B-rep, NURBS, parametric/structured CAD output)
    …click to see all
    MG

    Matheus Gadelha

    medium hireability

    Research Scientist@Adobe

    Previously: Research Assistant @ University of Massachusetts Amherst

    Seattle, US

    57
    Geometric Deep Learning82
    Generative Model Architectures75
    Generative 3D CAD72
    Structured Output Generation62
    Computational Geometry / CAD50
    RL for Generative Models3
    Strengths
    Proc3D (2026): PCG structured editable 3D generation with parametric LLM editing
    GEM3D (CVPR 2024): topology-aware generative 3D shape synthesis
    Gaps
    No B-rep, NURBS, or STEP file (CAD-specific geometry) expertise in publications
    …click to see all
    NM

    Nigel J. W. Morris

    medium hireability

    Autodesk

    Previously: Researcher @ Autodesk

    72
    Computational Geometry / CAD95
    Generative 3D CAD92
    Structured Output Generation85
    Geometric Deep Learning82
    Generative Model Architectures75
    RL for Generative Models0
    Strengths
    SolidGen (TMLR 2023): autoregressive B-rep synthesis — exact match to Spectral's SGS-2
    UV-Net (CVPR 2021): learned B-rep representations using UV parametric domains + topology graphs
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) in any published work
    …click to see all
    ND

    Nishkrit Desai

    medium hireability

    Researcher@Axiom

    Previously: Intern @ NVIDIA

    70
    Generative 3D CAD90
    Computational Geometry / CAD82
    Structured Output Generation82
    Generative Model Architectures82
    Geometric Deep Learning80
    RL for Generative Models5
    Strengths
    SolidGen (TMLR 2023, 89 citations): autoregressive B-rep synthesis
    ML techniques for direct B-rep synthesis (2024): continued CAD ML focus
    Gaps
    No RL/RLHF/DPO experience found
    …click to see all
    PJ

    Pradeep Kumar Jayaraman

    medium hireability
    90
    Generative 3D CAD98
    Computational Geometry / CAD95
    Generative Model Architectures95
    Structured Output Generation92
    Geometric Deep Learning90
    RL for Generative Models70
    Strengths
    BrepGen (SIGGRAPH 2024): diffusion model for structured B-rep generation
    SolidGen (2022): autoregressive B-rep synthesis, first-author lead
    Gaps
    Long Autodesk Research tenure (~5+ yrs) — no explicit move signals
    …click to see all
    RG

    Ruiqi Gao

    medium hireability

    Staff Research Scientist@DeepMind

    Previously: Research Scientist @ Google

    San Francisco, US

    31
    Generative Model Architectures85
    Geometric Deep Learning50
    Generative 3D CAD22
    Structured Output Generation18
    RL for Generative Models5
    Computational Geometry / CAD5
    Strengths
    CAT3D: multi-view diffusion for 3D generation (NeurIPS 2024)
    Bolt3D: fast 3D scene generation from inputs (2025)
    Gaps
    No structured/parametric CAD experience (B-rep, feature trees, STEP)
    …click to see all
    WY

    Wang Yifan

    medium hireability

    Researcher@Adobe

    Previously: Postdoc @ Stanford University

    32
    Geometric Deep Learning75
    Generative Model Architectures55
    Computational Geometry / CAD30
    Generative 3D CAD15
    Structured Output Generation10
    RL for Generative Models5
    Strengths
    ETH IGL PhD under Sorkine-Hornung — top computational geometry lab globally
    Eurographics Best PhD Thesis 2023 + ETH Medal — recognized 3D geometry authority
    Gaps
    No parametric CAD experience — B-rep, NURBS, feature trees, STEP all absent
    …click to see all
    ZS

    Zifan Shi

    medium hireability

    research scientist@Adobe

    Previously: PhD student @ The Hong Kong University of Science and Technology

    37
    Generative Model Architectures82
    Geometric Deep Learning62
    Structured Output Generation45
    Generative 3D CAD22
    Computational Geometry / CAD6
    RL for Generative Models5
    Strengths
    GRM (ECCV 2024, 235 citations): large-scale Gaussian 3D reconstruction + generation
    DMV3D (ICLR 2024 Spotlight, 207 citations): diffusion-based multi-view 3D model
    Gaps
    No structured/parametric CAD experience — work is visual 3D (radiance fields, Gaussians)
    …click to see all
    AJ

    Alec Jacobson

    low hireability

    Researcher@Adobe

    Previously: Senior Research Scientist @ Adobe

    Toronto, CA

    40
    Geometric Deep Learning92
    Computational Geometry / CAD65
    Generative Model Architectures35
    Structured Output Generation25
    Generative 3D CAD20
    RL for Generative Models5
    Strengths
    Creator of libigl — de facto geometry processing library
    NGLOD (607 cites) — neural implicit 3D shapes for real-time rendering
    Gaps
    No parametric CAD, B-rep, or STEP file work — misses core CAD representation requirement
    …click to see all
    AA

    Ali Mahdavi Amiri

    low hireability

    Vice President Research@Monsters Aliens Robots Zombies

    Previously: Research Director @ Monsters Aliens Robots Zombies

    North Vancouver, CA

    61
    Geometric Deep Learning85
    Computational Geometry / CAD78
    Generative 3D CAD70
    Structured Output Generation60
    Generative Model Architectures58
    RL for Generative Models12
    Strengths
    CAPRI-Net: compact CAD shape learning via primitive assembly (107 cit, CVPR 2022)
    D²CSG: CSG tree generation with dual complements (62 cit, NeurIPS 2023)
    Gaps
    No work on B-rep, NURBS, STEP, or parametric feature-tree generation
    …click to see all
    AT

    Andrea Tagliasacchi

    low hireability

    Associate Professor (Status Only)@University of Toronto

    Previously: Staff Research Scientist @ DeepMind

    Burnaby, CA

    44
    Geometric Deep Learning92
    Generative Model Architectures72
    Generative 3D CAD35
    Structured Output Generation33
    Computational Geometry / CAD28
    RL for Generative Models3
    Strengths
    BSP-Net: hierarchical structured mesh generation, 381 citations (CVPR 2020)
    Geometric deep learning core identity — SDFs, NeRF, implicit surfaces, point clouds
    Gaps
    No CAD-specific work: no B-rep, STEP, feature trees, or parametric solid modeling
    …click to see all
    AV

    Arash Vahdat

    low hireability

    Research Director@NVIDIA

    Previously: Senior Research Manager @ NVIDIA

    San Francisco, US

    31
    Generative Model Architectures90
    Geometric Deep Learning40
    Generative 3D CAD30
    Structured Output Generation15
    RL for Generative Models5
    Computational Geometry / CAD5
    Strengths
    LION (639 citations): latent point diffusion for 3D shape generation
    NVAE + Denoising Diffusion GANs — foundational diffusion architecture papers
    Gaps
    No structured/parametric CAD work — LION is point clouds, not B-rep or feature trees
    …click to see all
    CQ

    Charles R. Qi

    low hireability

    Member of Technical Staff@OpenAI

    Previously: Staff Machine Learning Scientist @ Tesla

    San Francisco, US

    38
    Geometric Deep Learning95
    Computational Geometry / CAD50
    Generative Model Architectures40
    Generative 3D CAD30
    Structured Output Generation10
    RL for Generative Models5
    Strengths
    PointNet/PointNet++ creator — foundational geometric DL on 3D point sets
    h-index 30; 21K+ citations on PointNet alone
    Gaps
    Career is 3D perception (detection/segmentation), not generative structured CAD
    …click to see all
    HS

    Hooman Shayani

    low hireability

    Principal Researcher@Autodesk

    London, GB

    68
    Geometric Deep Learning93
    Computational Geometry / CAD90
    Generative 3D CAD82
    Generative Model Architectures78
    Structured Output Generation55
    RL for Generative Models8
    Strengths
    BRepNet (CVPR 2021, 159 cites) — invented topological B-rep message passing
    UV-Net (CVPR 2021, 125 cites) — deep learning on boundary representations
    Gaps
    No RL/RLHF experience — clear gap for SGS-2 quality alignment work
    …click to see all
    JL

    Joe Lambourne

    low hireability
    80
    Generative 3D CAD99
    Computational Geometry / CAD98
    Geometric Deep Learning95
    Generative Model Architectures95
    Structured Output Generation90
    RL for Generative Models5
    Strengths
    BrepGen (SIGGRAPH 2024) — diffusion model for B-rep CAD generation
    SolidGen (TMLR 2023) — autoregressive direct B-rep synthesis
    Gaps
    No RL for generative models work (RLHF, DPO, GRPO) in any known publications
    …click to see all
    NM

    Niloy Mitra

    low hireability

    Principal Scientist@Adobe

    Previously: Researcher @ Adobe

    London, GB

    77
    Geometric Deep Learning95
    Generative 3D CAD92
    Computational Geometry / CAD90
    Generative Model Architectures88
    Structured Output Generation85
    RL for Generative Models10
    Strengths
    Free2CAD: freehand drawing → CAD commands (2022, 122 citations)
    SketchGen: constrained CAD sketch generation (2021, 109 citations)
    Gaps
    No evidence of RL for generative models (RLHF/DPO/GRPO)
    …click to see all
    PG

    Paul Guerrero

    low hireability

    Research Scientist@Adobe

    Previously: Research Associate (Post-Doc) @ University College London

    London, GB

    68
    Geometric Deep Learning88
    Generative 3D CAD85
    Generative Model Architectures80
    Structured Output Generation78
    Computational Geometry / CAD72
    RL for Generative Models5
    Strengths
    SketchGen (NeurIPS 2021) — constrained CAD sketch generation, exact domain match
    ShapeMOD + ShapeAssembly — structured 3D shape program synthesis
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) work evident
    …click to see all
    PJ

    Pradeep Kumar Jayaraman

    low hireability
    81
    Generative 3D CAD98
    Computational Geometry / CAD97
    Geometric Deep Learning93
    Generative Model Architectures90
    Structured Output Generation88
    RL for Generative Models20
    Strengths
    BrepGen (SIGGRAPH 2024): diffusion model for B-rep CAD — exact match
    SolidGen (2022) + AutoBrep (2025): autoregressive B-rep generation series
    Gaps
    No direct RLHF/DPO/GRPO work — RL-for-generative-models axis is weak
    …click to see all
    QC

    Qimin Chen

    low hireability

    PhD student@Simon Fraser University

    Previously: Research Scientist/Engineer Intern @ Adobe

    Burnaby, CA

    56
    Geometric Deep Learning82
    Generative 3D CAD72
    Computational Geometry / CAD63
    Generative Model Architectures62
    Structured Output Generation55
    RL for Generative Models3
    Strengths
    D2CSG (NeurIPS 2023): CSG tree decomposition from 3D shapes — CAD-native representation
    4 first-author SIGGRAPH-tier papers on 3D generative shape modeling
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) work found
    …click to see all
    RW

    Rundi Wu

    low hireability

    Research Scientist@DeepMind

    Previously: PHD Student @ Columbia University

    San Francisco, US

    73
    Generative 3D CAD95
    Generative Model Architectures88
    Structured Output Generation85
    Geometric Deep Learning80
    Computational Geometry / CAD80
    RL for Generative Models10
    Strengths
    DeepCAD (ICCV 2021) — lead author of foundational parametric CAD generation model
    PQ-NET (CVPR 2020) — sequential structured 3D shape generation (seq2seq)
    Gaps
    No RL/RLHF/DPO work in any published or public work
    …click to see all
    TG

    Thibault Groueix

    low hireability

    Research Scientist@Meta

    Previously: Research Scientist @ Adobe

    San Francisco, US

    43
    Geometric Deep Learning90
    Generative Model Architectures72
    Generative 3D CAD45
    Computational Geometry / CAD28
    Structured Output Generation18
    RL for Generative Models5
    Strengths
    AtlasNet (CVPR 2018, 1678 cites) — seminal 3D surface generation paper
    MeshUp (3DV 2025 Oral) — mesh deformation via blended score distillation
    Gaps
    No parametric CAD expertise — work is mesh/surface-based, not B-rep or feature trees
    …click to see all
    YF

    Yasutaka Furukawa

    low hireability

    Principal Scientist@Wayve

    Previously: Associate Professor @ Simon Fraser University

    77
    Generative 3D CAD97
    Generative Model Architectures95
    Structured Output Generation90
    Geometric Deep Learning88
    Computational Geometry / CAD85
    RL for Generative Models5
    Strengths
    BrepGen (SIGGRAPH 2024): B-rep generative diffusion model — most on-point paper for SGS-2
    SkexGen (ICML 2022): autoregressive CAD construction sequences with disentangled codebooks
    Gaps
    No RL for generative models (RLHF/DPO/GRPO) evidence in any publication
    …click to see all

    Runs

    #2completed0 qualified / 0 foundMay 5, 7:10 AM
    #1completed0 qualified / 0 foundMay 4, 5:51 AM