About Me

Ayushman Dash

I am a fast learner, hard worker, and a self-motivated independent developer. I have an affinity towards Artificial Intelligence (specifically Machine Learning and Deep Learning) and I aim at contributing as much as I can to these fields. My research interests are Sequence Learning, Generative Adversarial Networks, Sequence to Sequence Models, Siamese Network, Natural Language Understanding using Deep Learning, and Handwriting Recognition in Air. Apart from my interests in AI I have a keen interest in Web Application development, Music Composition and Production.

Ovation Summer Academy 2017 Berlin

Ovation Summer Academy 2017, Berlin

My Career

Element AI, Montreal, Canada

I bring state of the art Natural Language Processing research into products and custom solutions.

Applied Research Scientist

Insiders Technologies GmbH, Berlin

Currently, I am a member of the chatbot team at Insiders, and I am involved in developing their proprietary chatbot framework.

Oct 2018
Software Engineer - Natural Language Understanding

Ovation MindGarage

I am also working with MindGarage, an independent lab under the guidance of Prof. Marcus Liwicki, which with a vision of 'open minded research' focuses on Machine Learning and Deeplearning research. I help them in the administration, organisation and helping students in their projects.

Aug 2017
Student Assistant

Insiders Technologies GmbH, Kaiserslautern

I work in the Natural Language Understanding team and I am involved in developing a Deep Question Answering model for real-world business use cases. Apart from that, I have been involved in developing a Deep Neural Network component for fraud detection.

January 2017
Working Student

Deutsche Forschungszentrum für Künstliche Intelligenz GmbH (DFKI)

Now I am Working in the Multimedia Analysis and Data Mining team at Deutsche Forschungszentrum für Künstliche Intelligenz GmbH (DFKI) in the Multimedia Opinion Mining project. I am modelling trending events/news and analyzing them for visualizing opinions and trends in the world.

August 2016
Research Assistant

Department of Computer Graphics & HCI, TU Kaiserslautern

I worked on enhancing and maintaining the desktop client application for visualizing quality measures in automobiles for a project with Volkswagen.

Feb. 2016
Research Assistant

Technische Universität Kaiserslautern

I Joined the masters cource in Computer Science - Artificial Intelligence at Technische Universität Kaiserslautern.

Oct. 2015
Masters in Computer Science - Artificial Intelligence

Reverie Language Technologies Pvt. Ltd.

As a full stack web developer, I developed core products for Reverie and utility toolkits for internal teams. I Led a fast prototyping team of 3 for probable clients and developed and planned integration strategies with client applications. I also represented Reverie in tech events and competitions.

Feb. 2014
Full Stack Web Developer

Tata Consultancy Services Ltd.

As a Trainee I was trained on developing enterprise standard software applications and managing teams.

Sep. 2013
Software Engineer Trainee


On July 2013, I graduated from college and entered the real world.

Jul. 2013

My Skills











My Projects


Text Conditioned Auxiliary Classifier Generative Adversarial Network, (TAC-GAN) is a text to image Generative Adversarial Network (GAN) for synthesizing images from their text descriptions. TAC-GAN builds upon the AC-GAN by conditioning the generated images on a text description instead of on a class label. In the presented TAC-GAN model, the input vector of the Generative network is built based on a noise vector and another vector containing an embedded representation of the textual description. While the Discriminator is similar to that of the AC-GAN, it is also augmented to receive the text information as input before performing its classification.


Deep-Trans is a open-source project that I developed in collaboration with my ex-colleague Bhupen Chauhan. Considering the long temporal dependency of text in Indian languages and languages in general and with the domain Knowledge that we had from our past experiences, we extended our idea by creating a transliteration engine for English to Hindi phonetic conversion using an LSTM sequence to sequence model.


Pewter is an open-source project for data acquisition, analysis, visualization of raw data from Myo and conduct experiments on it. You can create experiments and visualize the data for doing some analysis before pre-processing and feature extraction. Pewter was originally developed by me for data acquisition and analysis of raw data from Myo Armband for one of my other projects, Voice.


LSTM.transpose() is an experiment with an unfolded version of LSTMs. The hypothesis is that The gradients of a deep Neural Network following the same architecture of the LSTM unfolded through time (even those of the bottom layers) are efficiently trainable with Backpropagation, and won't be affected by the 'vanishing gradient' problem. This is the case even when the weights are not 'tied'.


Air-Script is a CNN + Sequence to Sequence model for detecting handwriting on air using a Myo-Armband. It is Inspired by ‘Recursive Recurrent Nets with Attention Modeling for OCR in the Wild’ by Chen-Yu & Simon, 2016. The idea was to use 1D-CNNs as feature extractors and a sequence to sequence model with Attention mechanism introduced by Bahdanau et al., 2014 using LSTMs for variable length sequence classification.


A project for pretraining a Convolutional Autoencoder and then using the encoder weights to retraining another Convolutional Neural network as a transfer learning mechanism. It is being evaluated with the Tobacco dataset for documents.