Senior Applied Scientist, Amazon AGI Foundations
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Carnegie Mellon University
Master’s Degree, Artificial Intelligence
January 1, 2015 – January 1, 2016
Indian Institute of Technology (Indian School of Mines), Dhanbad
Integrated Master of Technology, Mathematics and Computing
January 1, 2008 – January 1, 2013
Sir Padampat Singhania Education Center
XII, Science, CS
January 1, 2006 – January 1, 2008
Amazon
Senior Applied Scientist
January 1, 2025 – Present
Amazon
Senior Applied Scientist
April 1, 2022 – Present
Amazon
Applied Scientist
February 1, 2018 – Present
Amazon
Applied Scientist
February 1, 2017 – January 1, 2018
Carnegie Mellon University
Graduate Teaching Assistant
August 1, 2016 – December 1, 2016
Greater Pittsburgh Area
Amazon
Applied Scientist Intern
May 1, 2016 – August 1, 2016
Greater Boston Area
Carnegie Mellon University
Graduate Student
August 1, 2015 – December 1, 2016
Greater Pittsburgh Area
Samsung Electronics
Software Engineer
June 1, 2013 – July 1, 2015
Noida Area, India
University of Calgary
Summer Research Assistant
June 1, 2011 – July 1, 2011
Calgary, Canada Area
Classifying real and fake text
April 1, 2016 – Present
Built a classifier to segregate real Broadcast news articles from trigram-generated fake articles. Used features such as n-grams, tf-idf, article length and perplexity w.r.t 3-gram and 4-gram models. SVM with linear kernel provided the best accuracy.
Re-ranking machine generated questions
January 1, 2016 – April 1, 2016
Built a generic question re-ranking module using Learning-To-Rank (LETOR) approach that can serve as last component of a question generation system. Experimented with length based, linguistic and language model based features. Used open source software SVM rank for training the system.
Question Generation and Answering System
January 1, 2016 – April 1, 2016
Worked in a team of 4 to build a system that can produce coherent, grammatical and fuent questions and answers. Used Wikipedia articles as dataset. The system was written in python using NLTK and Stanford CoreNLP tools and provides a command-line interface. Our team won the best project award.
Predicting star ratings from customer reviews
November 1, 2015 – Present
Built using python, a machine learning system from scratch to preprocess data, extract features (bag of words), and learn a multi-class classifier via logistic regression. Implemented stochastic gradient descent (SGH) and Batched SGH to train the logistic regression model. The work was done on a dataset from Yelp containing approximately 1 million reviews.
Analysis of fMRI data using Machine Learning
November 1, 2015 – December 1, 2015
Performed multi-class kernelized Support Vector Machine (SVM) classification to categorize the activity being performed by test subjects using their fMRI scans. In another task, used Principal Component Analysis (PCA) with ridge regression to predict missing voxels in fMRI scans of test subjects. Used holdout testing and cross-validation to test the model accuracy and prevent any over-fitting. Implementation was done in python using scikit-learn library.
Movie Recommender System
October 1, 2015 – Present
Developed, in MATLAB, a collaborative filtering based recommender system for movies using a subset of Netflix Prize dataset. Implemented both memory based and model-based approaches for rating prediction. Tried clustering techniques to address cold-start problem.
Text Based Information Retrieval
September 1, 2015 – December 1, 2015
Built Ranked Boolean, Okapi BM25 and Indri retrieval models in Java using Lucene search engine library. Implemented several query operators and added query expansion and learning-to-rank capabilities. Conducted experiments to analyze the performance of the search engine on ClueWeb09 dataset using metrics like Mean Average Precision (MAP) and Precision@N (P@N).
Clustering on News Articles
September 1, 2015 – Present
Build a Bipartite Clustering system, a type of reinforcement clustering that produces both document clusters and word clusters simultaneously, in MATLAB. Performed experiments on a subset of TDT4 dataset containing around 2000 documents. Evaluated the system's performance using Sum Of Cosine Similarity metric.
Fuzzy Transform and its Applications
May 1, 2012 – May 1, 2013
Studied and worked on Fuzzy Transform for discrete and continuous functions in one and two variables. Implemented and tested in Matlab, an approach to solve ordinary differential equations (ODE) with ordinary and fuzzy initial conditions using fuzzy transform. Also implemented in MATLAB, an image compression and reconstruction algorithm using fuzzy transform.
Proofs of Retrievability (POR) in Cloud Storage
June 1, 2011 – July 1, 2011
Implemented a cryptographic scheme (proof of retrievability) that had been proposed for verifiability of data storage in cloud computing environments. The implementation is in C++ and uses several open source libraries like GNU Multiple Precision Arithmetic library, OpenSSL and Pairing-Based Cryptography library.
Cultural Fit Analysis
The candidate's project portfolio demonstrates a strong inclination towards research and development in AI/ML, which aligns well with roles requiring innovation and deep analytical thinking. The diversity of projects, from text classification to fMRI data analysis and cryptographic schemes, indicates a broad intellectual curiosity and adaptability. The long tenure at Amazon as an Applied Scientist suggests a fit for large, technically driven organizations. However, the target role of 'Data Analyst' might be a slight mismatch for someone with extensive 'Applied Scientist' experience, potentially indicating an overqualification or a shift in career focus that needs further exploration.
Soft Skills & Operational Fit
The candidate's project descriptions indicate a strong problem-solving aptitude and a structured approach to complex data challenges. Participation in a team project (Question Generation and Answering System) suggests collaboration skills. However, without psychometric or English test results, a comprehensive assessment of communication, stress handling, and team collaboration is not possible.