^2000 Hello World!

I'm Niladri Shekhar Dutt

Machine Learning researcher || Computer Vision aficionado || NLP enthusiast


ARTificially Intelligent

I'm fascinated by AI and I like building architectures that exhibit “intelligence” and deploy novel models to solve real world problems. My research interests lie in the field of Deep Learning and its applications in Computer Vision and NLP.

Updates :

  • Will be speaking at PyCon Taiwan'19 on 'Probabilistic Programming using TensorFlow Probability' (September 21, 2019) - link
  • Won the 'Best Social Hack' at LA Hacks 2019 hosted by UCLA (Major League HAcking)
  • Won the 2nd Runners-up award at San Francisco (SF) Hacks 2019 (Major League Hacking)
  • Won the 'Best Utility Hack' at DeveloperWeek Hackathon, 2019 (America's largest challenge driven hackathon)

  • When I don't write code, I write snippets.


    January 2019 - June 2019

    University of California, Berkeley

    Visiting Student

    July 2016 - Present

    SRM University

    B.Tech in Computer Science Engineering

    My Work Experience.

    July 2019 - Present

    Data-X Lab, UC Berkeley

    Data Scientist

    Working with Dr. Ikhlaq Sidhu and Dr. Alexander Fred-Ojala on anomaly detection in car and aviation engines in a production pipeline for CMD Engines (Italy).

    January 2019 - May 2019

    CITRIS Lab, UC Berkeley

    Undergradute Researcher

    Working with Dr. Brandie Nonnecke, Director of CITRIS Policy Lab at UC Berkeley on large scale topic modeling in Twitter based on social network analysis and clustering techniques to improve topic coherence by 33-70%.

    June 2019 - June 2019

    IIT Kharagpur

    Research Intern

    Worked on autonomous driving research using CARLA simulator. Used computer vision and sequential modeling approaches based on Conditional Affordance Learning and a novel subset selection algorithm to achieve state of the art results on limited data.

    February 2019 - May 2019

    Data Science @ Berkeley

    Analytics Committee Member

    DSS@B is an undergraduate reserach club at UC Berkeley focusing on Data Science. I worked with on their data connector service on AWS.

    June 2018 - July 2018

    Cogknit Semantics

    Computer Vision Intern

    Worked on a computer vision powered fashion recommendation to suggest clothes which go along with each other with respect to trend, colour, pattern, fit, etc.

    January 2018 - Present

    Next Tech Lab

    AI Researcher

    We research , hack and acheive at Next Tech Lab, a QS award winning student-led research lab. I work on multidisciplinary research and deploy novel models to solve real world problems related to Computer Vision and Sequence modeling.


    Wasting time better

    Here's some of my favourite projects

    Self Driving Car simulation

    Conditional Affordance Learning and Subset selection

    Project Link

    Direct perception approach which maps video input to intermediate representations suitable for autonomous navigation in complex urban environments given high-level directional inputs. Designed a novel subject selection algorithm to drastically reduce the amount of data needed for traning while maintaing the same level of accuracy.

    Make a bet with your friend on the Ethreum Blockchain

    GitHub Link

    Won the overall 3rd prize at San Francisco (SF) Hacks (MLH)
    Easily create smart contract for a bet between 2 parties with a wager. The bet along with the wager which can be a product from Macy's will get deployed on the Ethereum blockchain. Use Betcy's Alexa skill to create a smart contract using just your voice. This is the most convinient way to create a smart contract.

    Brevis News

    Human like news summary and read multiple perspectives

    Project Link

    Uses a fine tuned GPT-2 to generate abstractive summaries which feel natural to the human eye. Also, provides multiple perspectives to a news article to reduce bias towards a particular political ideology. Entire pipeline built on : TensorFlow, React, AWS Lambda and Firebase


    Detect fake news

    Project Link

    NLP based model to detect fake news. The model achieved an accuracy of about 93%. We also created an algorithm to take in search queries as input and tell how true the search query is. The model was deployed using Django.

    Automated machine learning platform

    Project Link

    Selected for Y Combinator Online Startup School 2018
    A machine learning platform where data structuring, data preprocessing, feature engineering, and data modeling is automated. It makes usage of machine learning models a one click process. It also creates a detailed report of the data including visualizations. It has been deployed on the internet and is free to use.

    Emoji Translator

    Convert your text into emojis!

    Project Link

    Converts text to emojis in a fun way. There are two separate models to convert emojis to text:

  • Word to emoji - Convert each and every word to an emoji if it crosses a certain threshold using 300d GloVe vectors.
  • Sentence to emoji - Add an emoji at the end of the sentence to convey your sentiment. The model uses two stacked LSTMS and 100d GloVe vectors.