Aditya Khandelwal’s Resume

Big Tech

  1. Microsoft Shopping

    Pioneered GPT3-powered Smart-Compare Proof of Concept in Bing Chat, reimagining product comparison; scaled data pipeline to ~10M products/day in a resource-constrained setting; collaborated with key senior stakeholders for successful deployment

    Streamlined processing of an ML-powered Aspect-based Sentiment Analysis system by consolidating 5 pipelines across 4 global teams and 2 organizations, resulting in enhanced efficiency (10x processing time reduction) and accuracy (1.8x improvement)

    Led cross-functional effort to expand Product Graph taxonomy (novel way to store ~4B online product listings) by 70x, and show additional filters using it, resulting in ~9% revenue increase for affected listings, and high praise from CVP of Microsoft WebXT

    Developed an Active Learning-based fully automated ML system to find articles with shopping content, boosting Daily Active Users by 5x (200k to 1M); deployed in coordination with relevant organizations, ensuring minimal overlap with existing systems

Startup Experiences

  1. 2WheelR - Founding Team Member

    Helped with finances and growth strategy

  2. The Impact Engine - Founding Team Member

    Helped develop the product, set organizational strategy and execute strategic pivot

Publications

  1. NegBERT: A Transfer Learning Approach for Negation Detection and Scope Resolution

    Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5739–5748

    Authors: Aditya Khandelwal, Suraj Sawant

  2. Resolving the Scope of Speculation and Negation using Transformer-Based Architectures

    Authors: Benita Kathleen Britto, Aditya Khandelwal

  3. Multitask Learning of Negation and Speculation using Transformers

    LOUHI 2020: The 11th International Workshop on Health Text Mining and Information Analysis

    Authors: Aditya Khandelwal, Benita Kathleen Britto

  4. Orthogonal Attention: A Cloze-Style Approach to Negation Scope Resolution

    Creating a new Cloze-Style Attention mechanism inspired by Self Attention, and using it to address Negation Scope Resolution.

    Authors: Aditya Khandelwal, Vahida Attar

    Code (for 1 - 3): Transformers for Negation and Speculation

    Google Scholar Profile

Projects Completed

  1. Algorithmic API Generation for SimpleFix

    Used an XML file containing the structure of FIX Protocol messages to generate code for a higher-level API over SimpleFix.

  2. My DL Library

    Implemented basic Deep Learning concepts from Scratch using NumPy : (ANN, CNN, Optimizers like SGD, AdaGrad and Adam).

  3. Twitter Sentiment Analysis

    A deep learning model to implement Sentiment Analysis by a character-level Convolutional Neural Network, based on the paper by Xiang Zhang, Yann LeCun and Junbo Zhao. Model is scaled down to decrease training time and memory usage.

  4. Cryptron Game Playing Bot

    A bot that plays a game called Cryptron (Xodia’s problem statement for Credenz 2017). Uses MiniMax implemented with Alpha-Beta pruning, Iterative Deepening and a Principal-Variation Table.

Machine Learning Blog

Authors: Aditya Khandelwal, Renu Khandelwal

  1. Walk-through of a Machine Learning Example: The Iris Dataset

Lectures/Talks Given

Induction Lecture Series for COEP’s DS-AI Club

Video Playlist (YouTube)

  1. An Introduction into Python
  2. Functions and Objects in Python
  3. Important Libraries in Python (for Machine Learning)
  4. Machine Learning - 1
  5. Linear Regression from Scratch

Connect with me on LinkedIn