Pushpendra Pratap

About Me

I currently work in Bengaluru at BrightMoney. I like writing code and solving computational problems.

I can be reached by email at pratappushpendra3@gmail.com, and you may be interested in my official resume.

Languages & Tools

Languages: Python, SQL, Javascript, Unix shell scripting
Frameworks: Django, PySpark, Airflow, Pytorch, Scikit-learn, Flask
Databases: PostgreSQL, Neo4j, MongoDB, Redis
Libraries: Celery, RabbitMQ, AWS, Pandas, Numpy, Spacy, D3.js

Professional Experience

SDE-II (Data Platforms) at BrightMoney from September 2021 to Present

Built large scale maintainable system (for both real-time and offline use cases) which enabled our exponential growth. More specifically:

  • Built a very robust and low maintenance system to fetch and aggregate user data (e.g., accounts, transactions, etc.) from multiple sources (e.g., Plaid, Teller, Fiserv, Capital-one, Finicity, credit-report, etc.) both in realtime (i.e., during user onboarding) and offline (batch refresh jobs).

    Tech stack: Python, Django, PostgreSQL, Airflow, Celery, RabbitMQ, AWS, ELK stack, Grafana, Prometheus

  • Build a highly performant underwriting (UW) service (both for real-time and offline risk profiling). Realtime UW allows lots of feature-based split experimentation and is used for giving smaller loans/credits in less than 70 seconds (p99 latency) during user onboarding. Offline UW is for running on batches of users and supports bigger loans.

    Tech stack: Python, Django, PostgreSQL, Airflow, PySpark, Celery, RabbitMQ, AWS, ELK stack, Grafana, Prometheus

  • Optimized batch refresh jobs, made it horizontally scalable, reduced job time by almost 150%, and improved account recency by 30%.

  • Helped in transitioning from monolith to microservice-based architecture. Owned 6 different microservices.

  • Helped in database migrations.

  • Helped create the infrastructure that enabled large-scale data processing on hundreds of gigabytes of data.

  • Impact:

    • Increased total users from 30K+ to 240K+
    • Increased daily enrolled users from 400+ to 2500+
    • Reduced CPFC (cost for first charge) from $90 to $22
    • NOTE: We were able to achieve these numbers due to extraordinary contributions from other teams also.

Machine Learning Engineer at KiwiTech from September 2018 to September 2021

Designed and implemented novel solutions to tasks related to computer vision, NLP, backend, and MLOps.

  • Built the entire ML backend for a real-time action recognition app (achieves 94% video-level accuracy and 91% clip-level accuracy on the test set) along with the video annotation data pipeline for basketball games. Also implemented various filters (effects) that can be applied to different objects in the video, using object segmentation and tracking.

    Tech stack: Python, Pytorch, TensorFlow Lite, TorchServe, AWS Sagemaker, Pandas, Numpy, OpenCV, Docker, Celery, Redis

  • Designed the overall architecture and Built a web crawler to scrap thousands of websites, periodically (used TOR client to generate new IP addresses, to avoid blocking of crawler), and use those data to create a knowledge graph.

    Tech stack: Python, Django, PostgreSQL, Neo4j, MongoDB, Docker, Celery, RabbitMQ, Selenium

  • Built a web service to periodically fetch GitHub data & generate different visualizations (for monitoring purposes) of each GitHub repo (PR reviews, commits, committer, authors, etc.) of our organization.

    Tech stack: Python, Javascript, Django, ReactJS, d3.js, PostgreSQL, Docker, Celery, RabbitMQ

Machine Learning Engineer at Phonon.io from June 2017 to September 2018

Implemented various backend services and built NLP pipeline from scratch. More specifically:

  • Developed and deployed ML models (for tasks related to Intent Classification and Named Entity Recognition) for our clients like Aditya Birla Finance Limited (ABFL), which was used by over half a million users.

    Tech stack: Python, Flask, Scikit-learn, TensorFlow, Gensim, Spacy, Pandas, Numpy

  • Integrated different web APIs. Built logging API for Speech Recognition System. Built the underlying predictive model for the IVR system (for our client Max Life Insurance) and implemented various web services.

    Tech stack: Python, JavaScript, Django, Java, Hibernate, AngularJS


B.Tech in Computer Science from Jaypee University of Engineering & Technology(JUET) in 2017

Hackathons & Awards

May 2021

Airbus Aerothon 3.0 finalist

(270 candidates out of ~14000)