Ayush Pandey, Developer in Munich, Bavaria, Germany
Ayush is available for hire
Hire Ayush

Ayush Pandey

Verified Expert  in Engineering

Bio

Ayush is a deep learning engineer specializing in algorithms and APIs to enhance, analyze, and integrate complex projects. He worked as a data scientist at Outbrain, implementing and analyzing real-time machine algorithms, integrating them with the bidder infrastructure, and improving conversion rates by 30% in programmatic advertising. Ayush excels in projects that are highly mathematical and that require precision.

Portfolio

EventRegistry
Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT)...
Outbrain
A/B Testing, Data Analysis, Google BigQuery, BigQuery, Spark ML...
Sprinklr
REST APIs, Python, Spring, Java

Experience

  • Data Science - 5 years
  • Python - 5 years
  • Machine Learning - 5 years
  • Natural Language Processing (NLP) - 3 years
  • Generative Pre-trained Transformers (GPT) - 3 years
  • Keras - 3 years
  • Deep Learning - 2 years
  • TensorFlow - 2 years

Availability

Part-time

Preferred Environment

PyTorch, TensorFlow, Keras, Data Science, Julia, Python, MacOS, Windows, Linux

The most amazing...

...thing I built was the entire data science portion of a beta product in eight months at EventRegistry.

Work Experience

Deep Learning Engineer

2019 - PRESENT
EventRegistry
  • Developed a data science pipeline for a new product using deep learning-based text classification models.
  • Deployed the models using a REST API and performed A/B testing.
  • Wrote Python jobs to extract organizations, products, and locations from Wikidata.
Technologies: Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Flask, Text Classification, Deep Learning, Python

Data Scientist

2018 - 2019
Outbrain
  • Improved the conversion rate by 30% in programmatic advertising using a cutting-edge machine learning model.
  • Deployed A/B testing to production; analyzed and generated deep insights to drive up the conversion rate.
  • Researched and demonstrated the usefulness of Apache Spark as a replacement for Pandas for CTR analysis on a 300 GB data set.
Technologies: A/B Testing, Data Analysis, Google BigQuery, BigQuery, Spark ML, Amazon S3 (AWS S3), Go, Advertising Technology (Adtech), Machine Learning, Python

Data Engineer

2017 - 2018
Sprinklr
  • Served as product engineer to implement the machine learning data pipeline for real-time or batch processing using Kafka, Spark, and Spring framework.
  • Implemented a text classification engine using deep learning algorithms as the project's data scientist.
  • Deployed the models for clients using a REST API.
Technologies: REST APIs, Python, Spring, Java

Experience

Classification of Qualitative Traits of Differential Equation Solutions Using Machine Learning Tools

https://ayush-iitkgp.github.io/posts/gsoc-2017/
Implemented weighted non-linear regression, regularization, two-stage, and multiple shooting techniques to estimate the parameters of differential equations, which has its applications in HIV-AIDS viral dynamics study. Wrote an API to estimate the probability distribution of the parameters using Stan without having to write the Stan code.

Support for Optimization with Complex Numbers in Convex.jl

https://ayush-iitkgp.github.io/stories/juliacon-2016-talk/
Implemented support for optimization with complex variables in Convex.jl, making it the first open-source package to support the above functionality with many applications in quantum physics and AC power circuit optimization. Wrote the sample tests, documentation, examples, and blog entries to encourage the optimization community to use Convex.jl.

Large Scale In-memory Real-time Analytics on Apache Spark Framework

Set up an Apache Spark and Hadoop (HDFS) cluster. Implemented and benchmarked the performance of Apache Spark and R machine learning algorithms such as logistics regression and decision tree on the dataset of up to 100 GB.

Novelty Detection Algorithms for Electricity Theft

https://ayush-iitkgp.github.io/posts/when-does-traditional-classification-algorithm-fail/
Used one-class classification algorithms to predict the users most likely to resort to electricity theft, given their information. Achieved a relative performance improvement of 20% over the traditional binary classification algorithms.

Education

2012 - 2017

Master's Degree in Mathematics and Computer Science

Indian Institute of Technology Kharagpur - Kharagpur, India

Skills

Libraries/APIs

Keras, TensorFlow, Spark ML, REST APIs, PyTorch

Tools

Git, BigQuery

Languages

Python, Julia, R, Java, Go

Frameworks

Spring, Flask

Platforms

Linux, Windows, MacOS

Storage

Amazon S3 (AWS S3)

Other

Data Science, Mathematics, Machine Learning, Text Classification, Statistics, Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Software Development, Version Control, Mathematical Modeling, Advertising Technology (Adtech), Deep Learning, Optimization, Data Analysis, A/B Testing, Google BigQuery

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring