Hi, my name is
Pratyush Pandey.
I build things for the web.
Pratyush Pandey 🙋♂️ is a software developer 🚀 and designer from India 🇮🇳 who occasionally speaks in the third person.
After three years of crafting corners of the Web for others and twelve years after blogs died, he knew it was time to build his own.
He’s written more about himself, but would rather hear from you.
About Me
Hello! I'm Pratyush Pandey, a passionate software developer from India. I believe good software should be invisible at best, magic and delightful to use at worst, and that simplicity is worth the extra effort to achieve. Simplicity is the whole point.
While I have written software for Fortune 500 companies and have received scholarships and awards by renowned institutions for my work, I don't expect you to be impressed by that. I certainly wouldn't be.
Currently, I am pursuing the final year of my Bachelors at Indian Institute of Technology Delhi (IIT Delhi). I am from Mumbai, so its only natural I still refuse to call Delhi home. Outside college, I am into debates, journalism and music and hope to revisit these soon.
Here are a few technologies I've been working with recently:
- TypeScript, React
- Tensorflow
- Django, Flask
- PyTorch
- VueJS, NodeJS
- PostgreSQL, MongoDB
Where I've Worked
Software Engineer II @ Google
January 2022 - Present
- L4/L7 Load Balancing Team in Google Cloud Networking (Arcus)
Software Developer @ NK Securities Research Private Limited
July 2021 - January 2022
- Implementing and Optimising post-trade analysis for generating business metrics
- Automating existing processes and maintaining useful tools for monitoring live market prices and strategy performance metrics
- Expanding Angel Investment arm of the firm, including analysis of startup and businesses to make investment decisions
Software Engineer Intern @ Microsoft
May - July 2020
- Piloted a feature in Word-Online Codebase to integrate a personalised, Intelligent Ribbon Tab in Word Web (Adaptive Ux)
- Successfully delivered a Pilot for the 'For You' ribbon tab, which is fully user customisable and predicts future commands based on user history.
- Worked with TypeScript, React, Redux and C# for development; and KQL for data analysis.
- Trained, Hosted and Integrated a Tensorow, Keras based ML model in the ribbon for prediction of future commands.
- Developed a feature to detect Context in the document (text/image/table) for more relevant future command predictions.
Software & Data Science Intern @ Bending Spoons
Nov 2019 - Jan 2020
- In my time at Bending Spoons, I worked to understand the behavior and measure the value of millions of users of our apps.
- Worked with internal support tools teams on App Store Optimisation (ASO). Worked with Google BigQuery and Google Cloud Platform to prioritise our Apps in locale based search results on the App Store.
- Created a MongoDB and Flask based RESTful web application to view and edit App metadata to improve results.
- Handled asynchronous updates using Celery, and integrated the tool with Slack using the Slack API.
Deep Learning Research @ SUTD
May - July 2019
- Worked as a research assistant in Singapore University of Technology and Design (SUTD), in the ISTD pillar in Dr Ernest Chong's research group.
- Generalised the Dropout Regularisation method of Deep Neural Networks to a novel two model training approach called "MergeConnect"; enhanced generalisation over unseen data by 2% over dropouts/connect.
- Outperformed dropout on state of art ResNet32, wide ResNet & ResNeXT network architectures using benchmarking datasets SVHN, CIFAR-10 and CIFAR-100.
- Tested method on ResNeXt with skip-connection, data-augmentation & momentum to achieve 96.4% accuracy on CIFAR-10 dataset, improving error rate over Dropouts by 0.8%.
Machine Learning Intern @ Siemens AG
May - July 2018
- Worked in Energy Management - Strategy division in Siemens, Mumbai. Received a Letter of Recommendation for work.
- Created ML classification model for predicting the time of breakdown of Siemens motors, based on its operational parameters (vibrations, temperature, rotation speed, etc) at the time of production.
- Used Naive-Bayes & Random Forest algorithms for classification to achieve 90.2% accuracy on the training data.
- Held Training sessions on the FutureLand focus topics to sensitize 400 Workers and Executives from various departments.
Software Engineer Intern @ Renegade Insurance (Covered By Sage)
July 2020 - Dec 2020
- Helped ideate the system design and requirements; Employed Microservices and SOLID architecture using Springboot & Java
- Used Docker and Kubernetes for containerization and deployment; Used Kafka with Azure OCR for tracking COI Compliance
- Used React & TypeScript and schema services for FE; MongoDB for database, AWS S3, Textract for OCR, CDN and storage.
- Delivered MVP within 2 months to get first mover’s advantage over domain competitor Thimble, leading to growth of Sage’s business
Some Things I've Built
Featured Project
Fifa Database Portal
A real time player rating, club affliation, and official FIFA stat app for European league players. Includes W/L prediction and future standings for current season, along with an option to bet on and create custom clubs, teams, lineups and players for fun.
- PostgreSQL
- Tensorflow
- Javscript
- React
- Redux
Featured Project
Maeve: A Cannon Playing AI Bot
Meet Maeve, an AI bot for Cannon (two-player abstract strategy game). Maeve uses Reinforcement TD Learning & search to beat (most) humans at the game. Written from scratch in C++, she uses many of the same strategies that enabled IBM's Deep Blue to beat Garry Kasparov at chess in 1997.
- TD Learning
- IDDFS Search
- α-β Pruning
- Minimax
Featured Project
Apocalypse: A survival game
A survival game made on Unity 3D for Windows/Linux/Mac platforms. The objective is simple - Dodge the enemies, earn points, collect power-ups, and most importantly, stay alive. All scenes, sprites etc were digitally designed from scratch.
- Unity 3D
- C#
- Adobe Illustrator
Other Noteworthy Projects
view the archiveParallel Matrix Multiplication using MPI
Developed a parallel approach to efficiently multiply two dense matrices using Message Passing Interface (MPI) using Blocking P2P, Collective, and Non-Blocking P2P protocols.
What's Next?
Get In Touch
I'm currently looking for full time opportunities post graduation starting Fall 2021. If you think this is a good fit or just want to say hi, do write in!
CV review requests are welcome too 😃