Kafka Essential Training

TensorFlow in Practice Specialization

Machine Learning

Deep Learning Specialization

Data Engineering Nanodegree

Bachelor's degree

Sub-Same Day, Amazon Delivery Experience (DEX)

Software Development Engineer

Helped students/professionals prepare for tech interviews

Mentor

Worked in Amazon's A9 org
Search Science and AI

- Worked on migrating ML workflows to Native AWS, enabling automated scalability based on workload demands. This also improved the logging and troubleshooting capabilities.
- Developed a customized batch workflow plugin for an external team to help them reduce human labelling cost of their experiments and datasets. This was achieved by auto labelling high confidence records using our ML models.
- Developed an API service that empowered external teams to access predictions and data analysis on their datasets using our ML models. This service allowed users to submit jobs, triggering multi-task batch workflows that encompassed feature generation, model predictions, and in-depth analysis.
- Worked on automating dataset creation and GBDT model training process to reduce the manual effort involved in training new ML models.
- Trained GBDT and NLP Transformer classifier models for different marketplaces to improve the search relevance score.

T

- Built a visualization tool to group contextually related infrastructure alerts (issues) to reduce the Mean Time to Resolution. Modeled the infrastructure dependencies as a graph problem and used graph algorithms like BFS, Union-Find to show the visualization and identify the root cause for a bunch of alerts.
- Developed a Machine Learning powered solution to predict the likelihood of a production deployment resulting in an emergency reversion.

Technologies Used:
Front-end Development: ReactJS, Redux, Angular, d3
Back-end Development: Python, Flask
Data Streaming: Kafka
Databases: DB2
Machine Learning: scikit-learn

Technology Associate

Added support for Netflix’s VMAF repository inside FFmpeg by first converting VMAF into a library making it callable from any C / C++ program and then developing a video filter in FFmpeg that can run VMAF and output vmaf score for two video files.

Video Multi-Method Assessment Fusion (VMAF) (https://github.com/Netflix/vmaf) is an algorithm that predicts the subjective quality of a video compared to a reference by combining multiple elementary metrics using SVM Algorithm.

FFmpeg (https://github.com/FFmpeg/FFmpeg) is a collection of libraries and tools to process multimedia content such as audio, video, subtitles and related metadata.

Student Developer at FFmpeg

Author & Founder

Designed and developed a cross-platform mobile app for both website domain owners and providers to help them set reminders about the expiry date of the TLS/SSL certificate(s).

Used PhoneGap framework (Apache Cordova), Java, JavaScript, IntelliJ.

Intern

Ad Cloud Data Systems

- Led the migration of Hive and Presto jobs from Qubole to AWS EMR, enhancing availability and significantly reducing operational costs.
- Reduced the cost involved in running custom reports service by more than 80% by devising an automated system that identified and disabled reports with no usage or empty data.
- Worked on migrating some of the existing Amazon SWF workflows to Argo. 
- Led a cost-saving initiative by identifying unused AWS resources and establishing S3 bucket expiration policies, leading to an annual cost reduction exceeding $50,000 in AWS expenditures.

Technologies: AWS, EC2, S3, EMR, Hive, Presto, Qubole, Kafka, Druid, Zookeeper, MySQL, Kubernetes, Docker, Bazel

Computer Scientist

Content Creator

Author of AlgoMaster Newsletter | Prev @ Amazon

Know about Ashish Pratap Singh's work experience, education, bootcamps, certifications, and much more. Ashish Pratap Singh