Help & Supportcareers@gravityer.com
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Experienced NLP & ML Professional | Expert in GenAI, LLMs, and Low Resource ML Design
Senior data scientist and ML/NLP engineer with 10+ years of R&D experience in ML, NLP, NLU, ASR, conversational AI, information retrieval, and end-to-end ML systems ACCOMPLISHMENTS • Domain-independent Active Learning service for low-resource data Uniphore o Designed, researched, and deployed domain- and language-agnostic active learning framework for efficient annotation for low-resource conversational AI projects, leading to better annotated datasets and reduced annotation time and costs o Designed proposal after understanding project lifecycle of the company, conducted experiments on company datasets to establish viability of the framework, and led global, cross-divisional team to put it into production o Was able to reduce annotation costs and time of NER by an average of 60%, reducing overall time and cost of newer NER models • Named Entity Search and Correction of ASR transcription Samsung Research America o Successfully deployed NER search and correction framework as part of the bigger post-ASR improvements service o Developed state-of-the-art solution that would take multiple ASR output candidates and fix errors in named entities using graphs, traditional, and deep learning algorithms without adding significant latency to the ASR system o Reduced errors by 60% with minimal increase in latency, and filed a patent • End-to-end information retrieval and text classification framework American Family Insurance o Designed and deployed an end-to-end ML service for NLP for efficient annotation of large annotations and autonomous feature engineering and model training using active learning, pyspark, and elasticsearch o Increased usage of ML solutions across divisions in the company and reduced latency and redundancy in project deployment o Published a paper at an international conference • Claim dissatisfac
New York University
Master's Degree, Computer Science
September 1, 2013 – January 1, 2014
Clemson University
Master of Science - MS, Electrical and Electronics Engineering
July 1, 2012 – August 1, 2013
University of Mumbai
Bachelor of Engineering (BEng), Electrical and Electronics Engineering
January 1, 2007 – January 1, 2011
Self-Employed Contractor
Applied AI Scientist
September 1, 2025 – Present
Toronto, Ontario, Canada · Remote
Career Break
Bereavement
July 1, 2025 – September 1, 2025
SurveyMonkey
Sr Machine Learning Engineer II
February 1, 2024 – June 1, 2025
Ottawa, Ontario, Canada · Hybrid
salesDNA.ai
Principal AI Scientist
March 1, 2023 – November 1, 2023
Greater Seattle Area · Remote
Uniphore
Senior AI Scientist
September 1, 2022 – March 1, 2023
Greater Seattle Area · Remote
Uniphore
Senior NLP Scientist
August 1, 2021 – September 1, 2022
Greater Seattle Area · Remote
Samsung Research America
Senior Machine Learning Research Engineer
April 1, 2019 – August 1, 2021
San Francisco Bay Area · On-site
Flipkart
Data Scientist
May 1, 2018 – April 1, 2019
San Francisco Bay Area · On-site
American Family Insurance
Data Scientist I
February 1, 2015 – April 1, 2018
Madison, Wisconsin Area · On-site
3 Top, Inc.
Natural Language Processing Intern
June 1, 2014 – December 1, 2014
New York, United States · On-site
Active Learning for Token-level NLU
November 1, 2021 – Present
● Developed algorithm and framework for active learning for token-level NLU to reduce annotation time and costs across domains and languages ● Established data framework for easier, real-time data management ● Implemented the framework for entity recognition tasks across multiple languages with successful reduction of at least 50% annotation resources ● Led a global, cross-divisional team to convert framework into a service for the company
In-Capsule Classification for Bixby
May 1, 2020 – July 1, 2021
● Deployed capsule (virtual assistant domain)-based models for Bixby virtual assistant ● Optimized existing traditional machine learning architecture in Java for speed and accuracy
Named Entity Search and Correction for Bixby
July 1, 2019 – July 1, 2021
● Developed Named Entity Correction for the speech-to-text outputs generated by ASR system with low latency and high precision ● Developed real-time algorithms for named entity search for utterances spoken or typed by user on Bixby virtual assistant ● Developed machine learning pipelines in Python using traditional machine learning and deep learning algorithms and deployed them in Java programming environment ● Filed a patent for the algorithm
Aspect-based sentiment analysis of reviews
June 1, 2018 – April 1, 2019
● Developed feature-based sentiment analysis model for user-entered review on Flipkart website. ● Active learning and democratic co-training algorithms used for annotation
Question Answering
May 1, 2018 – April 1, 2019
Developing NLP model using deep learning to generate answers for a product-based question entered by a user on Flipkart website. The model would use information stored in knowledge base, which contains information of the products. Developed question classification model that would classify whether the given question is answerable as yes/no, from the knowledge base, would require more information, or not answerable at all. Developed an entailment model using deep learning to determine whether answer to the given question is yes, no, or not answerable
Review Classification
May 1, 2018 – September 1, 2018
Developed a classification model to classify reviews written on Flipkart website. The model handled misspelled and unknown words, and returned sentence-based class distribution for the given review. Semi-supervised techniques (democratic co-training and active learning) were used to label the unlabeled dataset
Chatbot
February 1, 2018 – April 1, 2018
Worked on a NLP-based chatbot for American Family Insurance. Developed knowledge graphs from unlabeled, unstructured datasets
Topic Discovery for Catastrophic Feedback from Client
November 1, 2017 – March 1, 2018
Determine the main topic on which the customer gave feedback on for the given claim. Used clustering techniques to determine number of topics and topic for each of the clusters.
A Deep Learning Approach to Burst Detection
February 1, 2017 – May 1, 2017
Developed deep convolutional architectures to classify whether the given signal from a rodent’s brain is a burst (activity) or not. Keras and tensorflow were used.
Sentiment Analysis for Article Comments
December 1, 2016 – March 1, 2017
Created a model for supervised classification of initially-unlabeled comments into three classes using NLP and machine learning algorithms. Used RoCKET for labeling and evaluating dataset, and Python for feature engineering and classification of comments
Search Engines
September 1, 2016 – October 1, 2016
Developed NLP-based search engines for use-cases, such as searching through legal-based documents and project taxonomies, in Python. Elasticsearch and StanfordNLP were used to develop a AngularJ-based web UI to search and display results
Concept Identification for BI Claim Notes
May 1, 2016 – August 1, 2016
Wrote algorithms to implement RoCKET to label and classify a large number of claim notes generated during Bodily Injury Claims into twelve concepts, which were used as features for determining the severity of the given BI claim
RoCKET
February 1, 2016 – May 1, 2017
Stands for Robust Concept and Knowledge Extraction from Text. Developed towards a pipeline that encompasses the entire NLP pipeline, ranging from efficient labelling for huge, unlabeled datasets to feature engineering and classification. Used active learning, big data, NEXT, and Elasticsearch for developing pre-classification pipeline, and Python-based NLP libraries for classification model
Claim Dissatisfaction Prediction
February 1, 2015 – June 1, 2015
Developed machine learning models in Python for generating features from text-based data provided by the employees and customers of the company related to an insurance claim.
Dependency Parser for natural language processing
September 1, 2014 – December 1, 2014
Developing codes in Python to establish dependencies between the word elements of the categories entered by a user on the 3Top website using research papers and grammar rules.
Similar Category Linking
August 1, 2014 – December 1, 2014
Generated similarity measure among the categories entered by different user in order to incorporate the items entered by him/her under the same category
Named Entity Recognition and Linking
June 1, 2014 – December 1, 2014
Developing codes in Python to recognise various named entities in the categories entered by a user on the 3Top website, ranging from personal names to locations and names of books, television shows, and movies. Various APIs and datasets, such as GoogleMap, Yahoo GeoPlanet, and DBPedia, are used to recognise the names after applying PoS Tagging algorithm on the categories.
Part-of-Speech Tagging and Optimisation for natural language processing
June 1, 2014 – December 1, 2014
Developing and improving a part-of-speech tagging for categories entered by a user on 3Top website using pre-existing libraries in Python, such as NLTK, StanfordNLP, and TextBlob, as well as grammar rules.
Handwritten Character Recognition using Deep Learning
March 1, 2014 – May 1, 2014
Developed codes in Python to perform handwritten character recognition using a large amount of training data, using deep learning algorithms. The project implements the deep method version of LeNet convolutional networks
Social Networking Website for hobbyists
January 1, 2014 – May 1, 2014
Developed front-end and back-end of a social website, which aims at connecting hobbyists, using MySQL and PHP. The website will allow the users to upload multimedia files, make diary entries, establish friendships between people, and allow the user to post comments.
Face Recognition using Wavelets
January 1, 2014 – May 1, 2014
Developed an algorithm in MATLAB to achieve face recognition in the images using wavelets.
Handwritten Character Recognition using Deep Learning
January 1, 2014 – May 1, 2014
Developed codes in Python to perform handwritten character recognition using a large amount of training data, using deep learning algorithms. The project implements the deep method version of LeNet convolutional networks.
Email Summarisation
September 1, 2013 – December 1, 2013
Developed an algorithm to summarise email threads using local and global TF-IDF scores of nouns occurring in the threads. The project was programmed in JAVA using Eclipse IDE and GATE libraries.
Handwritten Character Recognition
January 1, 2013 – May 1, 2013
Developed a program that takes images of handwritten English text and saves characters recognised into a notepad file. Coding was done in MATLAB.
AR Drone as UAV testbed
September 1, 2012 – June 1, 2013
A supposedly thesis project abandoned due to change in university, the goal was to develop an algorithm to achieve autonomous gutter-following AR Drone 2.0 (developed by Parrot) using images. Had developed line-following algorithm for autonomous flight of the AR Drone before abandoning the thesis.
Surface and Underwater Locomotion and Communication
July 1, 2010 – April 1, 2011
Developed an amphibian wireless robot to move on land and over and under the surface of water.
Watermarking using Haar Wavelets
January 1, 2010 – June 1, 2010
The objective of the project was to understand and implement watermarking technology employed by various industries using Haar wavelet. This project dealt with the necessity of watermarking, different techniques that can be employed, and how the technique of Haar wavelets are effective in achieving the desired goal.
Core Elasticsearch: Developer
Elastic
June 24, 2026 – Present
Cultural Fit Analysis
The candidate's diverse project portfolio, ranging from academic research to industry applications in e-commerce, virtual assistants, and insurance, demonstrates adaptability and a broad interest in applying ML across different domains. Their experience in leading global teams and mentoring colleagues suggests a collaborative mindset. The target role of ML Engineer aligns well with their extensive background in ML/NLP research and development, indicating a strong cultural fit for an innovation-driven environment.
Soft Skills & Operational Fit
The candidate's project descriptions highlight leadership, mentorship, and cross-divisional collaboration, suggesting strong soft skills and operational fit for a senior role. Experience in leading AI divisions and establishing research-to-production pipelines indicates an ability to drive initiatives and manage complex projects. The career break for bereavement shows personal resilience, though it's not directly a soft skill.