Sagar Chaturvedi
ML EngineerSenior Applied Scientist, Amazon AGI
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Senior Applied Scientist, Amazon AGI
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
ML Tech Lead with 12+ years of cross-domain (ML/DL, NLP, LLMs, ResponsibleAI, Computer Vision, Information Retrieval, MLOps, Big Data) R&D experience building large scale ML based solutions that make decisions on millions of user's data everyday. Technically leading the design/implementation/deployment of end2end large scale ML systems, owning the engineering/infra and working with cross-functional teams of product managers, program managers and remote data annotation teams to define/derive/establish the customer impact through ML.
Rajasthan Technical University, Kota
Bachelor’s Degree, Computer Engineering
N/A – Present
University of Utah
Master’s Degree, Computer Science
N/A – Present
Amazon
Senior Applied Scientist
December 1, 2024 – Present
Bellevue, Washington, United States
Amazon
Senior Applied Scientist
November 1, 2019 – December 1, 2024
Bellevue, Washington, United States
American Family Insurance
Sr Machine Learning Scientist
June 1, 2018 – October 1, 2019
Greater Seattle Area
University of Utah
Researcher - Applied ML
January 1, 2017 – May 1, 2018
Salt Lake City Metropolitan Area
Scry Analytics
Data Scientist
August 1, 2015 – July 1, 2016
Indicus Analytics
Senior Software Engineer
June 1, 2015 – August 1, 2015
New Delhi Area, India
NEC Technologies India Ltd.
Senior Member of Technical Staff
July 1, 2013 – October 1, 2014
NEC Technologies India Ltd.
Member Of Technical Staff
July 1, 2011 – June 1, 2013
Matching Images with News Titles
January 1, 2017 – Present
- Given a news Images and some titles, find the most suitable title for the image. Discriminative Approach - - A Region-CNN to identify the regions and their features in the image - A bi-directional RNN (Tensorflow) to create a feature representation of the titles. - Created a multi-modal embedding combining the image region features and title features. - Identify the title with the best similarity with the image in the multi-modal embedding space. Generative Approach - - Created an Image captioning model based upon "Deep Visual Semantic Alignment" paper. -
Species Classification for Birds using Labeled Image Corpus
November 1, 2016 – Present
Trained multi-class classifiers to identify bird species from labeled data of 11800 Bird Images with 200 species. ● Experimented with 3 types of features - Handcrafted color and shape features (included with data), OpenCV SIFT features and features extracted using Image-net trained Convolutional neural net of Tensorflow. ● Experiment 1 - Handcrafted features (http://www.vision.caltech.edu/visipedia/CUB-200-2011.html) o ID3 decision tree using entropy and GINI measure gave 56% and 40% accuracies respectively o One-vs-Rest SVM and Logistic regression gave 31% and 32% respectively after parameter tuning o Pair-wise SVM and Logistic regression achieved up to 96 and 93% respectively after parameter tuning o SVM with non-linear RBF kernel achieved the best results - 67% for gamma = .01 and C = 10. ● Experiment 2 - SIFT features using OpenCV o Created clusters of SIFT keypoint features using bag-of-words approach with k-means clustering. o Experimented with k = 20, 100 and 1500 o Used clustering results as feature input to One-vs-Rest SVM but could not do better than 12% even after trying 36 combinations of the learning rate gamma and the hyper-parameter C ● Experiment 3 - Tensorflow CNN based features o Using SVM with RBF kernel, achieved the best accuracy of 87% on this dataset. o Pretty unstable results as it could only manage ~3% with One-vs-Rest SVM. o Trained 10 pair-wise classifiers by selecting two random labels at a time and achieved accuracies ranging from 49.153% to 100%.
Supervised ML Suite
October 1, 2016 – Present
This suite is being created for implementing, training, evaluating and testing some of the most important Supervised Machine Learning Algorithms - o ID3 decision tree with depth adjustment, entropy & GINI calculation and information gain calculation. o Margin perceptron, aggressive perceptron and standard perceptron with learning rate adjustment. o Random forests with adjustment in number of trees to be created. o Soft SVM with adjustment in learning rate and the hyper-parameter C. o Logistic regression with the adjustment in the regularization parameter.
Exploratory Data Analysis
Coursera
June 24, 2026 – Present
Practical Machine Learning
Coursera
June 24, 2026 – Present
R Programming
Coursera
June 24, 2026 – Present
Regression Models
Coursera
June 24, 2026 – Present
Cultural Fit Analysis
The candidate has a diverse background spanning large tech companies (Amazon), insurance (American Family Insurance), research institutions (University of Utah), and analytics firms (Scry Analytics, Indicus Analytics, NEC Technologies). This breadth of experience suggests adaptability and a willingness to work in various environments. Their involvement in Responsible AI and Trust/Safety initiatives aligns with a culture that values ethical considerations in technology. The personal projects demonstrate strong initiative and a passion for ML, indicating a proactive and self-driven individual.
Soft Skills & Operational Fit
The candidate's experience at Amazon as a Senior Applied Scientist, leading efforts in Responsible AI, LLM alignment, and defining science roadmaps, suggests strong leadership, problem-solving, and strategic planning skills. Their work on balancing latency goals with business metrics indicates an operational mindset. The detailed project descriptions imply good communication and analytical abilities.