Hi, I'm
Kartik Dutt.
I'm a Software Engineer at Microsoft.
Software Developer | ex - Zomato | GSoC Mentor | GSoC - 20 | DTU - 21 | Open Source Enthusiast
Projects.
Get Fit
A site to manage and track all workouts. Track your progress and build routines, you are just a click and some excercises away from getting fit. Why don't you try it?
- JavaScript
- React
- CSS
- HTML
- Git
- AWS
- Firebase
LeetCode CLI
A CLI interface to your favourite Data Structures and Algorithms website. Get questions by name or id, recomendations of questions and custom templates to test your solutions locally.
- Python
- GraphQL
- Git
Personalized Search Engine
Search through more than 1 million of web-pages in couple of microseconds using Golang based GUI.
- Golang
- Git
Activity Recognition
Built and trained a CNN-LSTM model for activity recognition on UCF Dataset.
- PyTorch
- Python
- Git
PyTorch-mlpack weight translator
Why should you train a resnet on the same task across different frameworks? Well, you should simply translate the model weights from one framework to another and save your resources. This projects helps in doing exactly that. Translate models like YOLO or Darknet written in PyTorch to mlpack in couple of commands.
- XML
- Python
- C++
Experience.
Software Engineer
@Microsoft
July 2021 — Present
Working with designer team (Office 365).
- C#
- React
Software Engineer
@Zomato
June 2021 — July 2021
Worked with Autosuggest team to develop experiments for Outlet selection logic, APIs for migration of outlet selection logic and added aditional tracking to various states in Autosuggest. Worked with Dish Search Team to implement rerankers for Dish Search Result Page as well as experiments on Previously Ordered Rails.
- Solr
- Go
- PHP
- Github
Software Engineer Intern
@Zomato
January 2021 — June 2021
Worked with Dish Search Team and contributed to Personalization in Dish Search Page by adding cron and worker for adding new ranking logic on Dish Search Page. Worked with Autosuggest team to develop internal dashboard for quick merchant query resolution, experiments to personalize orders for user, fixes and clean up in tracking and Autosuggest as well as debugged issues related to outlet visibility.
- Solr
- Go
- PHP
- Github
Software Developer
@mlpack, Google Summer of Code
May 2020 — September 2020
Implemented bench marking tools and object detection algorithms including YOLO-v3 and tiny versions for twice as fast training time. Restructured models repository by devising dataloaders and object oriented for models with unit testing with over 70% code coverage. Devised optimized metrics for object detection algorithms such as intersection over union and layers like mini-batch normalization which provides over 5 times the speed up on 1000 data points over traditional implementation in C++.
- mlpack
- Tensorflow
- PyTorch
- YAML
- Azure
- CMake
- C++
- Python
- Github
Machine Learning Intern
@Coding Blocks
May 2019 — October 2017
I worked on processing and analyzing data to create state of the art models in python using PyTorch framework with 80% validation accuracy. Increased accuracy of traditional deep learning models by 12% in breast cancer detection in CT scans using augmentation and further by 3% by tuning hyper parameters.
- Python
- Azure
- Keras
- PyTorch
Open Source.
mlpack
mlpack is an intuitive, fast, and flexible C++ machine learning library with bindings to other languages like Python, Go and Julia. I worked on creating over 50 PRs for adding features, debugging, testing, refactoring and documenting the code base as well as integrating azure pipelines.
- C++
- Git
- Azure
- Boost
- Cereal
- Armadillo
Code-kii
Code-kii is an organization that aims to provide open source platforms for conducting technical coding interviews, code execution engines and code sharing platforms. Contributed Share-Code repository to provide a platform for live collaboration with multiple users.
- Javascript
- ReactJS
- HTML
- CSS
- MongoDB
- AWS
- Git
Achievements.
Infosys Digital Makeathon v2
3rd
2019
Won 3rd place in 2019 out of more than 20 colleges including Indian Institute of Technology. Developed an algorithm to predict traffic congestion factor and attained 88% validation accuracy set using LSTMs and real time tweet clustering. We used Object Detection Algorithm to detect accidents or other similar events that might affect traffic. This is really interesting, take a look at our README in the github.
- React
- Javascript
- Python
- Keras
- PyTorch
- scikit-learn
- Google API
- Github
Flipkart GridTech
Top 15
2019
Round 1 (Quiz) : Ranked 1st in DTU and 67 in India out 6000 participants. In round 2 and 3 obtained a rank in top 15 teams in object localization challenge with 90% accuracy on given data-set.
- PyTorch
- Tensorflow
- Python
- Github
Infosys Digital Makeathon v1
2nd
2018
Built an autonomous banking security system using object localization and event classification algorithms.
- React
- Javascript
- Python
- Go
- Kotlin
- Github
