CV

Ivan Puhachov's CV, covering AI engineering, computer graphics research, education, publications, and professional experience.

Contact Information

Name Ivan Puhachov
Professional Title AI Engineer
Location Montreal, QC, Canada
Website https://puhachov.xyz/

Professional Summary

AI engineer with a PhD in Computer Science. Builds LLMs, VLMs, and AI agents. Published five papers at SIGGRAPH, SIGGRAPH Asia, and Eurographics.

Experience

  • Oct 2025 - present

    Montreal, Canada

    AI Engineer
    Coverstar, a16z speedrun
    • Fine-tuned and deployed a ModernBERT moderation filter on AWS for 15 million monthly comments. Reduced p99.9 latency by 67% while maintaining throughput.
    • Fine-tuned a VLM for video moderation and improved frame sampling for blur and low light.
    • Built a context-aware LLM moderation agent. Automated warnings and sent bans and escalations for human review. Livestream blocks doubled, and mean response time fell by a reported 40%.
    • Built a customer support agent with custom MCP tools, RAG, PII screening, and human review. Covered 200+ daily ticket events and reduced estimated manual labor by 75%.
    • Built a voice and text agent with LiveKit, STT, TTS, tool calls, and RAG for user memory. Tested conversation flows and moderation in CI/CD.
    • Automated image-to-3D asset generation with Meshy AI, rigging, and diagnostic animation. Cut estimated time to a first draft by about 80%.
    • Advised on real-time, on-device avatar lip sync in Unity.
  • Oct 2021 - Feb 2024

    Montreal, Canada

    Research Engineer
    Huawei Research, Canada
    • Built research demos for generative motion, mesh deformation, scene reconstruction, and digital avatars using NeRF, GANs, diffusion, and Gaussian splatting.
    • Led first-author SIGGRAPH Asia 2023 research from synthetic Blender data and computer vision training through optimization, evaluation, and publication.
    • Mentored an intern on SIGGRAPH Asia 2024 research. Set up multi-GPU training and experiment tracking with PyTorch Lightning and Weights & Biases.

Publications

  • 2026
    Generative Cutout Animation
    Eurographics 2026

    Ivan Puhachov, Noam Aigerman, Thibault Groueix, Mikhail Bessmeltsev
    Generates editable 2.5D character rigs from text and skeletal motion using SDXL, LoRA, score distillation, and differentiable rendering.

  • 2024
    Neural Implicit Reduced Fluid Simulation
    SIGGRAPH Asia 2024

    Yuanyuan Tao, Ivan Puhachov, Derek Nowrouzezahrai, Paul Kry
    Uses a DeepSDF latent space for fluid simulation. Reports a 1,000× speedup for in-distribution simulations with comparable visual fidelity.

  • 2023
    Reconstruction of Machine-Made Shapes from Bitmap Sketches
    SIGGRAPH Asia 2023

    Ivan Puhachov, Cedric Martens, Paul G. Kry, Mikhail Bessmeltsev
    Reconstructs CAD-like 3D meshes from bitmap sketches using computer vision and patch-based geometric optimization.

  • 2022
    Stability-Aware Simplification of Curve Networks
    SIGGRAPH 2022

    William Neveu, Ivan Puhachov, Bernard Thomaszewski, Mikhail Bessmeltsev
    Selects curves that preserve structural stability under worst-case loads using a greedy approximation.

  • 2021
    Keypoint-Driven Line Drawing Vectorization via PolyVector Flow
    SIGGRAPH Asia 2021

    Ivan Puhachov, William Neveu, Edward Chien, Mikhail Bessmeltsev
    Converts bitmap line drawings into editable vector curves using keypoint detection, graph processing, and PolyVector Flow.

Skills

Programming Languages: Python, C++, JavaScript
AI and Machine Learning: LLMs, VLMs, AI agents, RAG, MCP, fine-tuning, computer vision, diffusion models, NeRF, Gaussian splatting
Frameworks: PyTorch, PyTorch Lightning, Hugging Face, NumPy, OpenCV, LiveKit, OpenGL, CGAL, libigl, Eigen
Infrastructure and Tools: AWS EC2, AWS Lambda, CloudWatch, Docker, GitHub Actions, Langfuse, Weights & Biases, Blender
Geometry and Graphics: differential geometry, mesh optimization, deformation, animation, vector fields, graph optimization

Education

  • Sep 2019 - Aug 2025

    Montreal, Canada

    PhD
    Université de Montréal
    Computer Science, DIRO, LIGUM
    • Research supervisor: Mikhail Bessmeltsev
  • Sep 2013 - Jun 2019

    Ukraine, Italy

    MSc (cum laude) in Mathematical Engineering and BSc in Mathematics
    University of L'Aquila & Kharkiv National University
    Mathematical Engineering and Mathematics
    • Joint MSc Programme Intermaths
    • Thesis: “Catacaustics of surfaces” (advisor: Alexander L. Yampolsky)