Grigor Nalbandyan, Developer in Munich, Germany
Grigor is available for hire
Hire Grigor

Grigor Nalbandyan

Verified Expert  in Engineering

Data Scientist and AI Developer

Location
Munich, Germany
Toptal Member Since
July 22, 2022

Grigor is a data scientist with over three years of industry experience in applied machine learning. He focuses on deep learning, computer vision, and natural language processing and has co-authored a paper that was accepted at the Institute of Electrical and Electronics Engineers (IEEE). Grigor is also completing his master's in data engineering and analytics at the Technical University of Munich in Germany.

Availability

Part-time

Preferred Environment

PyTorch, NumPy, Pandas, Scikit-learn, Transformers, Python 3, Git, Machine Learning, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Data Science, Artificial Intelligence (AI), Computer Vision, Deep Learning, Object Detection, Computer Vision Algorithms, JSON, CSV, BERT, Word2Vec, JSTransformers

The most amazing...

...project I've worked on is an unstructured document information extraction based on computer vision.

Work Experience

Machine Learning Specialist

2018 - PRESENT
WebbFontaine
  • Developed a tool for the extraction of key information from images.
  • Co-authored the paper: "Tokengrid: Towards More Efficient Data Extraction from Unstructured Documents," which was accepted at the IEEE.
  • Built an optical character recognition (OCR) engine for text detection and recognition from images.
  • Classified product descriptions into greater than 3,000 classes of harmonized system codes.
Technologies: PyTorch, Pandas, NumPy, Scikit-learn, Transformers, Python 3, Python, Computer Vision, Machine Learning, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Data Science, Artificial Intelligence (AI), Deep Learning, Convolutional Neural Networks (CNN), Object Detection, Computer Vision Algorithms, Text Classification, fastText, BERT, Word2Vec, JSTransformers

Key Information Extraction from Scanned Images

https://ieeexplore.ieee.org/document/9749071
The project included the development of an in-house OCR engine and a computer-vision-based model for necessary data extraction. For both parts, I took part in data annotation management, research, development, and making the models production-ready.

As a result, we wrote the paper: "Tokengrid: Towards More Efficient Data Extraction from Unstructured Documents," which was accepted at the IEEE.

Research of Pruning and Quantization Effect on CNN-Based Models

A research project to learn the state of efficient inference of CNN-based models.
This project focuses on the following:

• The latency vs. accuracy tradeoff when pruning a CNN model.
• The latency vs. accuracy tradeoff when quantizing a CNN model.
• The effect of hardware potential speed-up.
• Comparison of various frameworks: PyTorch, ONNX, OpenVino, NeuralMagic

Search Engine for Classification of Product Descriptions Into Harmonized System

An AI-based search engine that can classify a product description into more than 3000 possible classes. Currently, it is being used by customs offices of several countries. It supports English, French, and Arabic.

Research of Neural Architecture Search (NAS) Algorithms

Extensive research of the current state of Neural Architecture Search. Main approaches used: Reinforcement Learning-based, Evolution, and Gradient-based. I implemented a Gradient-based search using the Differentiable Architecture Sample method.

Libraries/APIs

PyTorch, NumPy, Pandas, Scikit-learn, Matplotlib

Tools

Jupyter, Plotly, Open Neural Network Exchange (ONNX), Git, OpenVINO

Storage

JSON

Languages

Python 3, Python

Paradigms

Data Science

Platforms

Docker

Other

CSV, Transformers, Computer Vision, Machine Learning, Natural Language Processing (NLP), Artificial Intelligence (AI), Deep Learning, Convolutional Neural Networks (CNN), Computer Vision Algorithms, Text Classification, Linear Regression, Logistic Regression, Decision Trees, Gradient Boosting, Data Visualization, Neural Networks, Recurrent Neural Networks (RNNs), Long Short-term Memory (LSTM), fastText, Quantization, Neural Network Pruning, BERT, Word2Vec, JSTransformers, Generative Pre-trained Transformers (GPT), Object Detection, Artificial Neural Networks (ANN)

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring