Ayush Kumar

About Me

I am a Machine Learning Scientist at Observe.AI, a YCombinator-2018 startup, working in the field of voice AI. Prior to that, I was employed at IPsoft Inc. as R&D Engineer, working in Episodic Memory team of Amelia - an enterprise cognitive agent. In my practical experience, I have primarily worked in Conversational AI researching and productionizing solutions catering to intent and entities classification, dialog management, sentiment analysis and understanding call anatomy.

Download CV

Natural Language Processing (NLP)

Text Classification, Language Modeling, Conversational AI

ML/DL Libraries: Tensorflow, Pytorch, Sklearn

LR, RF, CNN, LSTM/GRU, Transformers

Python: Pandas, Numpy, Matplotlib

Research + Technical Writing

Technical Expertise

Sentiment Analysis

Multimodal sentiment analysis using gated attention mechanism, Unsupervised aspect detection, aspect based sentiment, lexicon creation in resource scarce languages like Hindi, Bengali, Dutch, Turkish etc.

Intent and Entity Detection

Supervised, semi-supervised and joint-learning using contextual CNN/LSTM architectures with attention mechanisms to identify intents and entities/slots.

Text Classification

CNN, LSTM/GRU, Transformers based architectures utilizing word2vec, glove, ELMo, BERT based representations for classification, similarity detection, text retrieval.

Task Oriented Dialog Systems

Worked on retrieval based dialog systems that leverages sequential matching networks and contextual LSTMs to learn context, query and response triplets.

Out of Domain Detection

Handled out of domain inputs for dialog systems using cascaded hierarchical filters: a mixture of heuristic and classifiers.

Data and Annotations

Extensively worked on creating data guidelines and annotation strategies which forms the base for any production ML systems. Explored teacher-student framework and active learning for data augmentation.

Blog

Contextualized Word Representations

How to generate word vectors depending on the context that can cater to challenge of polysemy along with capturing information regarding its syntactic and semantic characteristics. Description/Analysis of a method as published in “Deep Contextualized Word Representations” in NAACL, 2018 by Peters et al.

Retrofitting Word Vectors to Semantic Lexicons

How to integrate information that comes from existing lexicons to word vectors? A method as published in “Retrofitting Word Vectors to Semantic Lexicons” in NAACL, 2015 by Faruqui et al.

Topic based textual summarization

A document can contain information regarding multiple topics viz., sports and politics. A simple summarization method would present this document in a concise format (single summary). How about a setup that can regard topics present in the document to generate summary considering that topic (topic based summary)? Descriptions of a research work as published in "Generating topic oriented summaries using neural network" in NAACL, 2018 by Krishna and Srinivasan.

Achievements and Recognitions

ICASSP Paper

Work on multimodal sentiment analysis titled: 'Gated Mechanism for Attention Based Multimodal Sentiment Analysis' accepted in the 45th International Conference on Acoustics, Speech and Signal Processing, Barcelona, Spain. ICASSP with h-index of 80 is one the highest impact conferences in signal processing.

DAAD WISE Scholarship

Received DAAD WISE 2015 Scholarship for internship at institutions of higher education and research institutes in Germany.

Best Bachelor Thesis Nomination

My undergraduate work on sentiment analysis titled 'Leveraging Semantics in Unlabeled Corpus for Sentiment Analysis' got nominated for the best Bachelor Thesis. A part of the work got accepted in the form of publication at COLING titled 'A Hybrid Deep Learning Architecture for Sentiment Analysis'.

Best of SemEval

Our submission for Aspect-Based Sentiment Analysis in International Workshop on Semantic Evaluation 2016, collocated with NAACL conference featured in 'Best of SemEval' segment.

Shri Kedar Nath Das Memorial Award

Received an award for outstanding all round performance, excellent organizational abilities and leadership qualities in Department of Computer Science and Engineering, IIT Patna.