Susanta
Biswas

I currently work as a Software Engineer at Microsoft Azure, working on highly scalable distributed systems.
I worked as an Applied AI Engineer at Cognizant, designing and developing solutions leveraging ML in fields like Computer Vision, NLP etc.

I completed my Bachelor's degree in Computer Science and Engineering from National Institute of Technology Durgapur, where I undertook research in Computational geometry.

I am broadly interested in Large scale Distributed Systems, Machine Learning, Computer Vision, Sequence models and their applications. I love exploring and learning about new interesting things that excite me.

Discover More






Susanta
Biswas

I currently work as a Software Engineer at Microsoft Azure, working on highly scalable distributed systems.
I worked as an Applied AI Engineer at Cognizant, designing and developing solutions leveraging ML in fields like Computer Vision, NLP etc.

I completed my Bachelor's degree in Computer Science and Engineering from National Institute of Technology Durgapur, where I undertook research in Computational geometry.

I am broadly interested in Large scale Distributed Systems, Machine Learning, Computer Vision, Sequence models and their applications. I love exploring and learning about new interesting things that excite me.

Discover More

Technical
Skills

I have experience with large scale distributed systems and have built features for public cloud like Azure that empowers users at a global scale.
I am experienced in designing and developing high performance machine learning / deep learning solutions with scalable server side web backend in fields like computer vision etc. I have sound knowledge of Data Structures and Algorithms and well versed with web technologies.

Programming Languages: Python, C# .NET, C++, C
Web Development Frameworks / Libraries: Flask, MySQL, MongoDB, Angular, Elastic Search
DevOps / Libraries / Tools: Kubernetes, Docker, Helm, Azure, git
Machine Learning Frameworks / Libraries / Tools: OpenCV, Scikit-Learn, TensorFlow, Keras
Miscellaneous: Distributed Systems, Microservices, Web Development, Computer Vision, Machine Learning, Deep learning, NLP

Experience
Timeline

Work Experience

  • Microsoft
    Software Engineer
    August, 2021 - Present

    Azure MySQL

    Part of Azure MySQL team, design and develop scalable backend components for highly distributed Azure MySQL service.

    • Engineering owner for a major enterprise business critical, customer facing feature.
      • Designed the architecture that spans across the Azure MySQL’s control plane and data plane. Collaborated with other Azure service teams for requirements gathering, coordination, design-discussions etc.
      • Designed and developed scalable asynchronous REST APIs, CLI commands etc.
      • Developed actors that handles the feature’s workflow and can handle variety of scenarios like High-Availability (HA), Failover, Read-Replicas, Point-In-Time-Restore etc.
    • Designed and developed an auto-healing, crash-resilient substrate framework for major customer facing features.
      • Provides crash-resilience for onboarded applications. E.g., MySQL server crash, no connectivity etc.
      • Provides real-time operations like pause/resume/cancel of operation, progress, speed tracking etc.
      • Handles all the deployment, security, resource governance, tracking, monitoring, and many other APIs.
    • Auto-healing, crash-resilient resource cleanup actors that clean up application resources like storage, metadata etc.
    • Designed and developed a scalable data transfer module that can scale multi GB/s speeds with support for real-time monitoring metrics like progress, copy speed etc.

  • Cognizant
    Sr. Applied AI Engineer
    September, 2018 - August, 2021

    Emerging Technologies Lab, Comms Media Technology

    Worked as full-stack ML Engineer, designed and developed server-side web backend and machine learning algorithms for Machine Learning / Deep learning-based applications.

    • Team lead (team of 4) for an ML based end-to-end automation pipeline of document editorial workflow increasing productivity with potential savings of 1700+Hrs person-hours yearly.
    • Team lead (team of 5) for a facial posture based proctoring application that can generate insights like user’s line of sight, focus region, emotion, eye tracking etc.
    • Team Lead (team of 3) for an object detection and tracking system designed for low-end hardware. Designed and developed the entire backend machine learning and server-side web architecture.
    • End to end development of an anomaly detection system for file monitoring for a top tier technology company. Lead the design and development of machine learning and server-side web architecture. Developed Angular UI components and performed Dev-Ops deployment activities.
    • Designed and developed algorithms for detecting non-stationary Advertisement boxes in News videos.
    • Lead a team of developers (team of 2) for developing a model for outage prediction in data centres.
    • Decreased video processing time by 8,000% by designing, developing algorithms and a new optimized application pipeline for an Object detection application.

    Retail and Hospitality

    • As part of digital e-commerce team, developed an emotion perceiving chatbot capable of offering coupons based on the mood of user.

  • Complex Networks Research Group(CNeRG), IIT Kharagpur
    Research Intern
    May, 2017 - June, 2017

    Worked in Natural Language Processing domain. Designed and developed different word level and character level language models for word prediction using n-gram probabilistic model.

Education

Recent
Projects

Facial keypoint regression

Detection of facial keypoints using 15 landmark points on human face. Uses a CNN with regression type output.

Realtime Facial Recognition

Realtime facial recognition using Siamese neural network. The model uses vector encodings from face for identifying users.

Word prediction using n-gram probabilistic model

Next word prediction using n-gram probabilistic model with various smoothing techniques.

Stackoverflow question tagger

Generates topic tags for Stackoverflow questions using machine learning. The model can predict appropriate topic tags for a wide variety of StackOverflow questions.

Image Summarization

Caption generation or image summarization for images using a CNN-LSTM network.

Parts of Speech Tagger

Part of speech tagger for english language using Bidirectional LSTM network. The model converts a sentence to POS tags.

Realtime Facial Emotion Analyzer

Realtime human emotion analysis from facial expressions using CNN. The realtime analyzer assigns suitable emoji for the current emotion of user.

Voice Assistant

A LSTM based voice assistant which can do tasks when it recognizes the user speaking the trigger word. Here the trigger word is 'activate'.

Web based facial authentication system

Web based facial log in/ Sign up website which uses facial recognition for user authentication. The model runs on the server side and accepts authentication requests from clients.

Data structures and Algorithms

Solution code for 400+ data structure and algorithm problems. The problems are partitioned topic-wise and written in C++14.

English to French Language Translator

Sequence to sequence model for Language translation from English to French using Neural Machine Translation(NMT). The model uses an Encoder-Decoder network with LSTM cells.

Name generation using RNN

Generates Indian names using Recurrent Neural Network (RNN). It uses a character level language model with Gated Recurrent Units for this.

Similar question finder

Find similar questions using word embeddings. It uses Google's 300 dimensional word embeddings.

Phrase to emoji convertor

Deep RNN for converting an english phrase to an emoji. Using a deep LSTM network the model associates an english input sentence with an emoji.

Neural Date translator

Date translation from conventional Human readable format to machine readable format(YYYY-MM-DD) using Neural Machine Translation(NMT). The model converts dates to a normalized format like YYYY-MM-DD.

Word Analogy

Finding word analogies using GLoVe word embeddings. In the word analogy task, we have 'a is to b as c is to __'. For example: 'boy is to girl as king is to queen'.

Classic snake game

Classic snake game written in C# using Windows forms.

Multitab Text Editor

Multitab Text Editor written in C# using Windows forms.

Live HTML editor

HTML editor that can show the live web view of the HTML code being written.

Code Snippet Manager

Code snippet manager written in C# that can be used to store and manage code snippets. Also has an in-built text editor to modify and write snippets.

Technical
Certifications

image 1
image 2
image 3
image 4
image 5
image 6
image 7
image 8
image 9
image 10
image 11
image 12