Senior Computer Vision Scientist · Bayer Crop Science

Seyed Vahid
Mirnezami

I build computer vision and deep learning systems that run in production, from multi-modal imaging and drone-based field analytics to generative AI and vision–language models. I also co-create LatentQuest courses, where I teach AI from the math up.

See my work LinkedIn
  • Ph.D.Mechanical Engineering, Iowa State University
  • 10+ yearsin machine learning and computer vision
  • 3patent filings as an inventor
Portrait of Seyed Vahid Mirnezami
Building vision–language systems

About

From plant phenotyping research to AI in production

I'm a Senior Computer Vision Scientist at Bayer Crop Science, where I work on deep learning, generative AI, and large vision–language models for real-world production systems. My work spans multi-modal imaging, drone and remote-sensing analytics, transformers for biological sequences, and the MLOps infrastructure that takes these models from notebooks to deployment.

Before Bayer, I was a data scientist intern at Evisort, building document-intelligence models for table and signature detection. I earned my Ph.D. at Iowa State University, majoring in Mechanical Engineering with minors in Computer Science and Plant Breeding, with research on high-throughput monitoring of plant growth using machine learning and deep learning.

Current focus

Generative AI and vision–language models for production, and automated deep learning pipelines with active learning on Amazon SageMaker.

Built in production

Multi-modal image models (RGB, X-ray, hyperspectral, fluorescence and laser), UAV-based plant counting, insect segmentation, and DNA-BERT for biological sequences.

Teaching

Co-creator of LatentQuest, teaching transformers, BERT, Vision Transformers and semantic search from first principles.

Experience

Where I've worked

  1. Senior Computer Vision Scientist

    Bayer Crop Science · St. Louis, MO

    Jul 2019 – Present
    • Developing deep learning, generative AI and vision–language models for production systems.
    • Building infrastructure that automates the deep learning pipeline (development, testing and deployment) with MLOps, integrating active learning and CI/CD on Amazon SageMaker.
    • Implemented DNA-BERT with transformers to encode DNA sequences, and validated its predictions of RNA-binding protein preferences for Arabidopsis annotations.
    • Developed a multi-task model that classifies and segments insects on leaves with over 0.92 accuracy.
    • Fine-tuned ResNet on multi-modal images (RGB, X-ray, hyperspectral, fluorescence and laser), and found the most informative channels with an LSTM attention model, saliency maps and feature selection.
    • Selected field sampling locations with machine learning and remote-sensing images, optimized drone flight paths, and designed an LSTM model to count germinated plants from UAV images.
    • Leading a cross-team project to define a standard label format for image datasets used by computer vision teams across the company (since 2021).
    • Mentored a computer vision intern who built a multi-task segmentation and classification model, and interviewed 30+ candidates for five roles.
    • Computer vision
    • Transformers
    • MLOps
    • SageMaker
    • Remote sensing

  2. Data Scientist Intern

    Evisort · San Francisco Bay Area, CA

    Oct 2018 – Apr 2019
    • Detected tables in documents with an autoencoder and recognized their content with OCR.
    • Detected signatures with object detection and classified signed documents.
    • Built a web application with Dash and Flask on Google Cloud Platform.
    • Worked with the data science team on NLP models for text feature extraction.
    • Document AI
    • Object detection
    • OCR
    • GCP

  3. Graduate Research Assistant

    Iowa State University · Ames, IA

    Aug 2015 – Jul 2019
    • Ph.D. research on high-throughput monitoring of plant growth using machine learning and deep learning.
    • Led two senior computer science students who built a web-based phenotyping application for plant breeders and biologists (2017 and 2018).
    • Plant phenotyping
    • Computer vision
    • Deep learning

Ph.D., Iowa State University

Major in Mechanical Engineering · Minors in Computer Science and Plant Breeding · 2015 – 2019

Teaching

LatentQuest courses

With Mohammad R. Khalili Shoja, I co-create courses that teach AI from the math up: real numbers first, then PyTorch, then real projects.

Available now

Master Transformers from Scratch: BERT, ViT & AI Search

From vectors and attention math to BERT, Vision Transformers and two-tower semantic search.

Enroll on Udemy Course details Get a coupon

Free code from the course

self-attention-from-scratchAttention with real numbers, in NumPy and PyTorch transformer-encoder-pytorchThe encoder block, and why positional encoding matters vision-transformer-cassavaFine-tune a Vision Transformer on Kaggle semantic-search-two-towerTwo-tower retrieval and re-ranking from scratch

Skills

Tools I work with

Machine learning

  • Deep learning
  • 2D & 3D computer vision
  • NLP
  • Transformers
  • BERT
  • Vision–language models
  • Generative AI
  • Feature engineering & selection

Frameworks

  • PyTorch
  • TensorFlow
  • Keras
  • OpenCV
  • scikit-learn
  • pandas
  • NetworkX
  • statsmodels
  • NLTK

Cloud & MLOps

  • AWS
  • Amazon SageMaker
  • GCP
  • Spark
  • CI/CD
  • MySQL
  • Linux

Languages & web

  • Python
  • C++
  • C#
  • R
  • JavaScript
  • MATLAB
  • Django
  • Flask
  • Dash
  • Plotly

Let's connect

Open to conversations about computer vision, vision–language models, and teaching AI. The fastest way to reach me is LinkedIn.

Message me on LinkedIn GitHub LatentQuest