“Have worked with Ayushman as team lead on multiple projects for a year. Highly motivated, dedicated, energetic and friendly colleague. Good at first thoughts, leadership skills, bright thinking and innovative ideas. When it comes to work, sheer quality reflects in his output of the work.”
Über uns
As the CEO of NeuralSpace, I am on a mission to bring cutting-edge AI research to our…
Activity
-
TiE Bhubaneswar warmly welcomes Mr. Ayushman Dash, CEO of NeuralSpace, as our newest Associate Member. With over a decade of expertise in natural…
TiE Bhubaneswar warmly welcomes Mr. Ayushman Dash, CEO of NeuralSpace, as our newest Associate Member. With over a decade of expertise in natural…
Liked by Ayushman Dash
-
Thrilled to share that we’ve signed a major international client: Brightcove one of the world’s largest video platforms. Brightcove has…
Thrilled to share that we’ve signed a major international client: Brightcove one of the world’s largest video platforms. Brightcove has…
Liked by Ayushman Dash
-
🌟 Presenting the Expert Voices at CIO100! We’re thrilled to showcase an exceptional lineup of expert speakers from our esteemed partner…
🌟 Presenting the Expert Voices at CIO100! We’re thrilled to showcase an exceptional lineup of expert speakers from our esteemed partner…
Liked by Ayushman Dash
Erleben Sie
Bildung
-
Technische Universität Kaiserslautern
-
Activities and Societies: Software Laboratory
I am doing my Masters in Computer Science and specialising in Artificial Intelligence. I have been an active member of the Software Laboratory at the Department of Computer Science at the University. I have been involved in many projects and activities as a part of the lab including the following,
Created a demo android application for Indoor Positioning System which can locate and navigate a user indoors.
Developing a machine learning based application for converting sign…I am doing my Masters in Computer Science and specialising in Artificial Intelligence. I have been an active member of the Software Laboratory at the Department of Computer Science at the University. I have been involved in many projects and activities as a part of the lab including the following,
Created a demo android application for Indoor Positioning System which can locate and navigate a user indoors.
Developing a machine learning based application for converting sign language to speech.
Worked as a Software Developer in The Visualization and HCI department of the University. -
-
Activities and Societies: Arpeggio (Music Club), College Band (Shabd), Ziarza (Robotics Club)
Founded the first music club in the University. Actively participated won various competitions in renowned Universities like IIT Kanpur, IIT Bhubaneswar, XIM Bhubaneswar as a part of the College Band.
Actively participated as a volunteer for Bakul in many events like IDOS (International Day of Service) and Annual day of OLS (Open Learning System, an institution for differently abled children).
Attended lessons in manual and automated robotics (in 1st semester) at Ziarza (College…Founded the first music club in the University. Actively participated won various competitions in renowned Universities like IIT Kanpur, IIT Bhubaneswar, XIM Bhubaneswar as a part of the College Band.
Actively participated as a volunteer for Bakul in many events like IDOS (International Day of Service) and Annual day of OLS (Open Learning System, an institution for differently abled children).
Attended lessons in manual and automated robotics (in 1st semester) at Ziarza (College Robotics Club).
Licenses & Certifications
Volunteer Experience
-
Event Coordinator and Performer
Bakul Foundation
- 3 Jahre
-
Student Volunteer
MindGarage
- Present 8 years
Science and Technology
The Ovation MindGarage is an independent lab under the guidance of Prof. Marcus Liwicki. It is connected to the teaching activities of Marcus in the area of deep learning, including the lecture "Very Deep Learning", his supervised projects and theses, and individual studies.
Publications
-
Subword Semantic Hashing for Intent Classification on Small Datasets
Institute of Electrical and Electronics Engineers (IEEE) Explore
In this paper, we introduce the use of Semantic Hashing as embedding for the task of Intent Classification and achieve state-of-the-art performance on three frequently used benchmarks. Intent Classification on a small dataset is a challenging task for data-hungry state-of-the-art Deep Learning based systems. Semantic Hashing is an attempt to overcome such a challenge and learn robust text classification. Current word embedding based methods are dependent on vocabularies. One of the major…
In this paper, we introduce the use of Semantic Hashing as embedding for the task of Intent Classification and achieve state-of-the-art performance on three frequently used benchmarks. Intent Classification on a small dataset is a challenging task for data-hungry state-of-the-art Deep Learning based systems. Semantic Hashing is an attempt to overcome such a challenge and learn robust text classification. Current word embedding based methods are dependent on vocabularies. One of the major drawbacks of such methods is out-of-vocabulary terms, especially when having small training datasets and using a wider vocabulary. This is the case in Intent Classification for chatbots, where typically small datasets are extracted from internet communication. Two problems arise with the use of internet communication. First, such datasets miss a lot of terms in the vocabulary to use word embeddings efficiently. Second, users frequently make spelling errors. Typically, the models for intent classification are not trained with spelling errors and it is difficult to think about ways in which users will make mistakes. Models depending on a word vocabulary will always face such issues. An ideal classifier should handle spelling errors inherently. With Semantic Hashing, we overcome these challenges and achieve state-of-the-art results on three datasets: Chatbot, Ask Ubuntu, and Web Applications. Our benchmarks are available online.
-
AirScript
The 14th IAPR International Conference on Document Analysis and Recognition
This paper presents a novel approach, called AirScript, for creating, recognizing and visualizing documents in air. We present a novel algorithm, called 2-DifViz, that converts the hand movements in air (captured by a Myo-armband worn by a user) into a sequence of x, y coordinates on a 2D Cartesian plane, and visualizes them on a canvas. Existing sensor-based approaches either do not provide visual feedback or represent the recognized characters using prefixed templates. In contrast, AirScript…
This paper presents a novel approach, called AirScript, for creating, recognizing and visualizing documents in air. We present a novel algorithm, called 2-DifViz, that converts the hand movements in air (captured by a Myo-armband worn by a user) into a sequence of x, y coordinates on a 2D Cartesian plane, and visualizes them on a canvas. Existing sensor-based approaches either do not provide visual feedback or represent the recognized characters using prefixed templates. In contrast, AirScript stands out by giving freedom of movement to the user, as well as by providing a real-time visual feedback of the written characters, making the interaction natural. AirScript provides a recognition module to predict the content of the document created in air. To do so, we present a novel approach based on deep learning, which uses the sensor data and the visualizations created by 2-DifViz. The recognition module consists of a Convolutional Neural Network (CNN) and two Gated Recurrent Unit (GRU) Networks. The output from these three networks is fused to get the final prediction about the characters written in air. AirScript can be used in highly sophisticated environments like a smart classroom, a smart factory or a smart laboratory, where it would enable people to annotate pieces of texts wherever they want without any reference surface. We have evaluated AirScript against various well-known learning models (HMM, KNN, SVM, etc.) on the data of 12 participants. Evaluation results show that the recognition module of AirScript largely outperforms all of these models by achieving an accuracy of 91.7% in a person independent evaluation and a 96.7% accuracy in a person dependent evaluation.
Other authorsSee publication -
TAC-GAN
Arxiv
In this work, we present the Text Conditioned Auxiliary Classifier Generative Adversarial Network, (TAC-GAN) a text to image Generative Adversarial Network (GAN) for synthesizing images from their text descriptions. Former approaches have tried to condition the generative process on the textual data; but allying it to the usage of class information, known to diversify the generated samples and improve their structural coherence, has not been explored. We trained the presented TAC-GAN model on…
In this work, we present the Text Conditioned Auxiliary Classifier Generative Adversarial Network, (TAC-GAN) a text to image Generative Adversarial Network (GAN) for synthesizing images from their text descriptions. Former approaches have tried to condition the generative process on the textual data; but allying it to the usage of class information, known to diversify the generated samples and improve their structural coherence, has not been explored. We trained the presented TAC-GAN model on the Oxford-102 dataset of flowers, and evaluated the discriminability of the generated images with Inception-Score, as well as their diversity using the Multi-Scale Structural Similarity Index (MS-SSIM). Our approach outperforms the state-of-the-art models, i.e., its inception score is 3.45, corresponding to a relative increase of 7.8% compared to the recently introduced StackGan. A comparison of the mean MS-SSIM scores of the training and generated samples per class shows that our approach is able to generate highly diverse images with an average MS-SSIM of 0.14 over all generated classes.
Courses
-
Applications of Artificial Intelligence
-
-
Automata Theory (Stanford - Coursera)
-
-
Cambridge ESOL (Business English)
-
-
Case Based Reasoning
-
-
Collaborative Intelligence
-
-
Embedded Intelligence
-
-
Graph Theory
-
-
Human Computer Interaction
-
-
Intelligent User Interface Design
-
-
Linguistics and language Processing
-
-
Middleware for Heterogeneous and Intelligent Systems
-
-
Modelling Real-World Problems as Graphs and Complex Networks
-
-
Multimedia Analysis and Data Mining
-
Projects
-
Autodoc
- Present
A project for pre-training a Convolutional Autoencoder and then using the encoder weights to retraining another Convolutional Neural network as a transfer learning mechanism. The idea is to use this for document classification and not use Alex net for feature extraction and present a comparative analysis.
-
DeepTrans
DeepTrans is a character level language model for transliterating English text into Hindi. It is based on Tensorflow's Sequence to Sequence models. It uses Deep Recurrent Neural Networks for transliterating English to Hindi.
Other creatorsSee project -
Pewter
Pewter is an open-source project for
1. Data acquisition
2. Analysis
3. Visualisation
of raw data from Myo and conduct experiments on it. If you are working on raw data from the Myo Armband, then you can make use of Pewter's simple GUI to acquire data and work on it. It not only reduces devlopement time but also makes the life of a machine learning engineer easier. You can create experiments and visualise the data for doing some analysis before preprocessing and feature…Pewter is an open-source project for
1. Data acquisition
2. Analysis
3. Visualisation
of raw data from Myo and conduct experiments on it. If you are working on raw data from the Myo Armband, then you can make use of Pewter's simple GUI to acquire data and work on it. It not only reduces devlopement time but also makes the life of a machine learning engineer easier. You can create experiments and visualise the data for doing some analysis before preprocessing and feature extraction. All the experiment data will be saved in Json format which makes it even convenient for reuse. The entire project is created using Node.js and I thank Thalmiclabs for the myo.js library which made my life a lot more simpler. -
Voice
Sign language to speech conversion for people with full and partial speech impairment.
Currently working towards the development of an application that detects sign language and converts it into speech in real-time using the following,
• Myo Armband
• MaryTTS (The MARY Text-to-Speech System by DFKI)
• Deep Neural Networks -
Localization Manager
A workflow / CMS application to manage the workflow of manual translations by the translation team. This application was integrated with Language as a Service for easier client specific content management for real-time localization and building a parallel corpus for translation.
-
Language as a Service
A Web Service based application for end to end localization (Transliteration, Translation, multilingual search, language support detection) of mobile and web applications.
Other creatorsSee project -
Machine Reasoning for Sentiment Analysis with N-Gram CNN and Attention Based Networks
-
Deep Learning has led to a paradigm shift in the field of Natural Language Processing
(NLP). Despite its success in NLP tasks, their "black box" nature restricts them
from explaining causality. NLP tasks, like sentiment analysis demand aspect or
reason extraction from a given input text. This work is an attempt to show a more
intuitive way to use Deep Learning methods for modelling and extracting reasons
from a text.
This thesis introduces two novel methods that can be used in…Deep Learning has led to a paradigm shift in the field of Natural Language Processing
(NLP). Despite its success in NLP tasks, their "black box" nature restricts them
from explaining causality. NLP tasks, like sentiment analysis demand aspect or
reason extraction from a given input text. This work is an attempt to show a more
intuitive way to use Deep Learning methods for modelling and extracting reasons
from a text.
This thesis introduces two novel methods that can be used in various NLP tasks.
(1) An N-Gram Convolution Neural Network (N-Gram CNN) architecture, that
extracts features form variable length phrases, and (2) an Attention based Machine
Reasoning technique to extract causality.
Sentiment analysis was the best fit for showing the validity of the proposed methods
as it demands reason extraction from the input text to explain the cause of the
predicted sentiment.
Both the N-Gram CNN and the Attention model produce results that are comparable
with the state of the art results on the Stanford Sentiment Treebank (SST) dataset
and the [41] IMDB movie reviews dataset [30] for sentiment analysis. As opposed
to the other models, the proposed models have been evaluated on two additional
datasets: (1) German Amazon reviews dataset; (2) Hotel reviews dataset.
An additional contribution of this thesis is a generalization of the proposed methods in the field of Reinforcement Learning, and reasoning tasks like Visual Question Answering. Although the proposed attention-based model gives a general idea (without any external supervision) about the causality, it still requires further extensions to filter out meaningless results and to produce a polarity for the extracted reason. -
Air-Script-Seq
-
Air-Script is a CNN + Sequence to Sequence with attention model for detecting handwriting on air using a Myo-Armband.
-
TAC-GAN
-
A Tensorflow implementation of the Text Conditioned Auxiliary Classifier Generative Adversarial Network for Generating Images from text descriptions
-
LSTM.transpose()
-
An experimental project to see how a transposed single layer LSTM cell unrolled over time like a deep multi-layer perceptron would help in learning interesting features.
Honors & Awards
-
Microsoft Code For Honor
Microsoft Corporation
As a part of Reverie, I was one of the three, including our CTO and one of my colleagues who represented our company at the Microsoft Code For Honor 2014. The competition was organised for companies to showcase their state of the art solutions in various categories like Government Enablement, Industry, Society etc that can make a difference. We won the 1st prize in the category of Government Enablement and had showcased our Language as a Service platform that included all kinds of services…
As a part of Reverie, I was one of the three, including our CTO and one of my colleagues who represented our company at the Microsoft Code For Honor 2014. The competition was organised for companies to showcase their state of the art solutions in various categories like Government Enablement, Industry, Society etc that can make a difference. We won the 1st prize in the category of Government Enablement and had showcased our Language as a Service platform that included all kinds of services including input method to web services for transliteration, reverse transliteration and localization for automatic localization of application content. It took about 6 months to go through the first round till the finals and I represented the company in all of them. During this process, I was involved in public speaking, technical interviews and solution architecture discussions.
-
Trelleborg Appstars
Trelleborg AG
Trelleborg Appstars was a competition for budding application developers to showcase their out of the box ideas in the form of mobile applications. We had developed an android application that acted as a modulator for a MIDI controller and had an x-y grid for allowing 2 degrees of freedom to a musician for modulation the audio signals getting generated by the MIDI controller. It was a superb gig and we not only ended up winning the competition but also playing an awesome show in front of the…
Trelleborg Appstars was a competition for budding application developers to showcase their out of the box ideas in the form of mobile applications. We had developed an android application that acted as a modulator for a MIDI controller and had an x-y grid for allowing 2 degrees of freedom to a musician for modulation the audio signals getting generated by the MIDI controller. It was a superb gig and we not only ended up winning the competition but also playing an awesome show in front of the audience. We were highly praised for our unique presentation skills and confidence with which we made our pitch.
Test Scores
-
GRE revised General Test
Score: 305
-
TOEFL iBT
Score: 102
Languages
-
Englisch
Full professional proficiency
-
Oriya
Native or bilingual proficiency
-
German
Elementary proficiency
-
Hindi
Native or bilingual proficiency
Recommendations received
2 people have recommended Ayushman
Join now to viewMore activity by Ayushman
-
Talent or discipline? Confused !⁉ A real haunting question to the recruiters 😖. We hired 2 guys in 2022 with almost 🏅Similar years of…
Talent or discipline? Confused !⁉ A real haunting question to the recruiters 😖. We hired 2 guys in 2022 with almost 🏅Similar years of…
Liked by Ayushman Dash
-
People that you look upto are also figuring it out & failing gloriously. Embrace it!
People that you look upto are also figuring it out & failing gloriously. Embrace it!
Liked by Ayushman Dash
-
Zomato Feeding India Concert (ft. Dua Lipa) tickets go live for everyone at noon today! Thank you for the fantastic response so far! Those of you…
Zomato Feeding India Concert (ft. Dua Lipa) tickets go live for everyone at noon today! Thank you for the fantastic response so far! Those of you…
Liked by Ayushman Dash
Other similar profiles
Gemeinsame Artikel erkunden
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Ayushman Dash in India
-
Ayushman Dash
Engineering Manager | Product Development | Scrum Master @ GEP Worldwide
-
Ayushman Dash
Human Resource Specialist at Fessorpro | Empowering Talent and Driving Growth
-
Ayushman Dash
Python | Statistics | SQL | Machine learning | Deep learning | NLP | Power BI | Tableau | AWS | Django | RestAPI | Docker
-
Ayushman Dash
Intern @Unacademy | Ex-SRE-Intern @Signeasy | Building MaYoTea | Founder @Enthalpy | VITCC-24
29 others named Ayushman Dash in India are on LinkedIn
See others named Ayushman Dash