Dilip Mathew Thomas, Developer in Kochi, Kerala, India
Dilip is available for hire
Hire Dilip

Dilip Mathew Thomas

Verified Expert  in Engineering

Machine Learning Developer

Location
Kochi, Kerala, India
Toptal Member Since
April 11, 2019

Along with earning a PhD in computer science and engineering, Dilip has over a decade of experience in the industry. Since 2015, he's been focusing on projects related to machine learning and deep learning. Dilip has an eye for detail which helps in working closely with domain scientists and improving the accuracy and reliability of models for fine-grained image classification, object detection and segmentation, natural language processing, time-series forecasting, and generative AI.

Portfolio

Independent Consultancy
Scikit-learn, PyTorch, Keras, Artificial Intelligence (AI)...
Vyby Inc
Artificial Intelligence (AI), Stable Diffusion, Google Cloud Platform (GCP)...
Photograde
Scikit-learn, Deep Learning, Computer Vision, Redis, FastAPI, Flask, REST APIs...

Experience

Availability

Part-time

Preferred Environment

Git, Scikit-learn, PyTorch, Keras, Ubuntu

The most amazing...

...project I've worked on was the automation of a factory using an array of cameras and computer vision techniques.

Work Experience

AI/ML Consultant

2015 - PRESENT
Independent Consultancy
  • Performed the role of CTO for several early-stage startups by translating high-level product requirements into technical requirements. Designed experiments and guided junior engineers to build and evaluate AI models.
  • Implemented academic publications and customized them per clients' requirements. Provided hands-on expertise in deep learning and machine libraries like PyTorch, TensorFlow, Keras, Hugging Face, and scikit-learn.
  • Architected NLP and stable diffusion models with the Hugging Face library.
  • Worked with domain experts to understand the nuances and biases in the data and used their feedback to create better features and data to train AI models, improving accuracy and reliability.
  • Built fine-grained visual classification models using a combination of classification and metric learning techniques for improved accuracy and robustness.
  • Constructed generative image models for image generation from a sketch by considering the user requirements for the style of the image.
  • Executed text recognition from images using convolutional recurrent neural networks.
  • Developed an object-detection model for apparel detection.
  • Created prototypes for anomaly detection in a surveillance camera video feed using unsupervised techniques.
  • Designed prototypes for the human crowd counting on video feeds from street cameras.
Technologies: Scikit-learn, PyTorch, Keras, Artificial Intelligence (AI), Natural Language Processing (NLP), Hugging Face, Fine-tuning, Graphics Processing Unit (GPU), Data Science, Computer Vision, Deep Learning, Machine Learning, Algorithms, Data Scientist, Linux, Python 2, Python 3, Git, Python, XGBoost, Pandas, Programming, Integration, User Interface (UI), Models, Cloud, OpenCV, AI Programming, Research, Convolutional Neural Networks (CNN), Image Processing, Image Analysis, Large Language Models (LLMs), Data Analysis, Generative AI

AI/ML Consultant | Generative AI

2023 - 2023
Vyby Inc
  • Converted business use cases to technical problem statements for building an MVP.
  • Generated images using Stable Diffusion based on text prompts created by ChatGPT.
  • Converted the generated image into a video rendering with 3D photography using context-aware layered depth in-painting.
  • Built back-end APIs using the Stable Diffusion AUTOMATIC1111 web UI's API functionality.
Technologies: Artificial Intelligence (AI), Stable Diffusion, Google Cloud Platform (GCP), Natural Language Processing (NLP), Hugging Face, Deep Learning, Machine Learning, Computer Vision, Linux, Python 2, Python 3, Git, Python, Graphics Processing Unit (GPU), Data Scientist, Pandas, Programming, Integration, Models, Cloud, OpenCV, AI Programming, Convolutional Neural Networks (CNN), Image Processing, Large Language Models (LLMs), Data Analysis, Generative AI

AI/ML Consultant | Model Accuracy Improvement & Deployment

2022 - 2023
Photograde
  • Improved ML models' accuracy to be at par or better than competitors by interacting with domain experts and designing and implementing experiments to select the right model training features.
  • Designed and implemented asynchronous APIs using FastAPI and a job management system using Redis and job queues to scale the model deployment for concurrent use by multiple users.
  • Deployed the model in production and developed back-end Flask REST API interfaces to interact with the model.
  • Developed back-end Flask REST API interfaces to train a model to learn a user's photo editing style.
Technologies: Scikit-learn, Deep Learning, Computer Vision, Redis, FastAPI, Flask, REST APIs, Google Cloud Platform (GCP), Python-rq, Machine Learning, Artificial Intelligence (AI), MySQL, Data Scientist, Linux, Python 2, Python 3, DevOps, Git, Python, Fine-tuning, Graphics Processing Unit (GPU), XGBoost, Pandas, Programming, Integration, User Interface (UI), Models, Cloud, OpenCV, AI Programming, Data Visualization, Convolutional Neural Networks (CNN), Image Processing, Image Analysis, Data Analysis

AI/ML Consultant | R&D Quantitative Finance

2020 - 2022
AlphaBeta
  • Implemented several academic research papers from scratch and trained and back-tested machine learning models like transformers, time series models, CNN models, random forest, and gradient boosting trees on finance data.
  • Developed code for computing various technical and fundamental financial variables.
  • Wrote code for extracting and post-processing market and earnings data from finance databases.
Technologies: Artificial Intelligence (AI), Machine Learning, Scikit-learn, TensorFlow, PyTorch, XGBoost, Keras, MySQL, Data Scientist, Linux, Python 2, Python 3, Git, Python, Fine-tuning, Graphics Processing Unit (GPU), Pandas, Programming, Models, AI Programming, Research, Data Visualization, Data Analysis

AI/ML Consultant | R&D Computer Vision

2017 - 2021
Streamoid Technologies
  • Developed highly accurate object detection models for fashion categories by correcting for data biases. Built object detection models for small objects and worked on optimizations to improve inference speed.
  • Improved the accuracy of fashion attribute classification models through careful analysis of model weaknesses, experimenting with better techniques, and eliminating data biases.
  • Generated pixel-level annotated data using traditional computer vision algorithms and trained semantic segmentation models. Improved the accuracy of semantic segmentation models by analyzing annotation errors and correcting them.
  • Designed and developed a custom color classification CNN network using a pixel-voting scheme to detect the dominant color in fashion apparel in highly noisy images.
Technologies: PyTorch, TensorFlow, Scikit-learn, Keras, Deep Learning, Computer Vision, Linux, Python 2, Python 3, Git, Python, Fine-tuning, Graphics Processing Unit (GPU), XGBoost, Data Scientist, Pandas, Programming, Integration, User Interface (UI), Models, Cloud, OpenCV, AI Programming, Research, Bitbucket, Data Visualization, Convolutional Neural Networks (CNN), Image Processing, Image Analysis, Data Analysis

AI/ML Consultant | R&D Computer Vision

2015 - 2018
Uncanny Vision
  • Created and implemented several experiments for developing an anomaly detection model for video surveillance use cases using autoencoder neural networks and one-class classification methods.
  • Improved the human pose detection model's accuracy by debugging the convergence issues during model training.
  • Developed an automatic number-plate recognition model for very challenging vehicle number-plate recognition scenarios. Used computer vision and computer graphics techniques to synthetically generate vehicle number-plate training data.
  • Designed and developed a system for reading analog meter values using object detection, segmentation, and number recognition models.
  • Trained models for recognizing text in images using CNN and LSTM models.
Technologies: Deep Learning, Artificial Intelligence (AI), Machine Learning, Python 3, Python, Caffe, TensorFlow, Keras, PyTorch, Scikit-learn, Computer Vision, Linux, Git, Fine-tuning, Graphics Processing Unit (GPU), Data Scientist, Programming, Models, Cloud, OpenCV, AI Programming, Research, Bitbucket, Data Visualization, Convolutional Neural Networks (CNN), Image Processing, Image Analysis, Data Analysis

Computer Scientist

2015 - 2015
Adobe
  • Investigated the use of topological methods for data analysis.
  • Explored research use cases for Adobe's digital marketing portfolio.
  • Designed a topic modeling system to understand user behavior and engagement from their mobile phone usage.
Technologies: Python, Data Science, Artificial Intelligence (AI), Computer Vision, Deep Learning, Machine Learning, Algorithms, Git, Scikit-learn, Data Scientist, Programming, Data Visualization

Technical Staff Member

2006 - 2007
NetApp
  • Designed and implemented a data deduplication module for a virtual tape library.
  • Developed a proof of concept to show the effectiveness of data deduplication.
  • Maintained the back-end code for a content management module.
Technologies: C++, C, Data Science, Artificial Intelligence (AI), Algorithms, Programming

Senior Software Engineer

2002 - 2006
Philips
  • Designed and implemented enhancements for the workflow management of cardiovascular intervention software.
  • Created and developed a memory management module for efficient image storage and retrieval.
  • Built an import-and-export module of a patient database.
  • Performed the onsite system integration and testing at Philips Medical Systems, Netherlands.
  • Produced test cases and tested different modules before the release of the software.
Technologies: C++, C, Algorithms, Programming

Anomaly Detection in a Surveillance Video Feed

This project reviewed different techniques for detecting anomalies in videos using unsupervised machine learning techniques. We explored the use of reconstruction, predictive, and deep learning-based generative models for this project.

These reconstruction-based models build representations that minimize the reconstruction error of training samples from the normal distribution. Spatio-temporal predictive models consider the correlation by viewing videos as a spatiotemporal time series and learning representations that minimize the prediction error on spatiotemporal sequences. The generative models learn to generate samples from the training distribution while minimizing the reconstruction error and the distance between generated and trained distribution. Each of these methods focuses on learning prior information useful for constructing the representation for the video anomaly detection task.
2009 - 2015

Ph.D. in Computer Science and Engineering

Indian Institute of Science - Bangalore, India

2007 - 2009

Master of Engineering Degree in Computer Science and Engineering

Indian Institute of Science - Bangalore, India

1998 - 2002

Bachelor of Technology Degree in Computer Science and Engineering

National Institute of Technology - Calicut, India

Libraries/APIs

Keras, Scikit-learn, Pandas, OpenCV, PyTorch, TensorFlow, VTK, REST APIs, XGBoost, Python-rq

Tools

Bitbucket, Git

Languages

Python, Python 3, Python 2, C++, C

Paradigms

Data Science, DevOps

Storage

Redis, MySQL

Platforms

Linux, Ubuntu, Google Cloud Platform (GCP)

Frameworks

Caffe, Flask

Other

Machine Learning, Deep Learning, Computer Vision, Artificial Intelligence (AI), Algorithms, Natural Language Processing (NLP), Fine-tuning, Graphics Processing Unit (GPU), Programming, Models, AI Programming, Research, Convolutional Neural Networks (CNN), Image Processing, Data Analysis, Computational Topology, Scientific Data Analysis, Hugging Face, Integration, Cloud, Data Visualization, Image Analysis, Large Language Models (LLMs), Computational Geometry, Stable Diffusion, FastAPI, Data Scientist, Machine Language, User Interface (UI), Generative AI

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