Ashutosh Pandey

Experience

• 

  Wipro

  May 2017 - Jul 2017

  Summer Intern

  Under the Turbo internship program, I did my summer internship on “Continuous Integration and Continuous Development framework for probe-based analytics solution for C-RAN” under the guidance of Mr. Subhrajyoti Panja at Kolkata.

• 

  Accenture

  Sept 2018 - Aug 2021

  - Built an ASR(Automatic Speech Recognition) custom model (6-8 months) based on 4000 Hours of rawdata using transfer learning. (Skills used: Data Analytics, SQL, Python, NLP, RESTful API, Nginx)- Designed a solution for Audio Transcription in python using multiple AWS Services (Major Skill used:AWS Transcribe, Comprehend, S3, EC2, Glacier, CodeCommit, Python, Restful API, Nginx, Tomcat,XML & JSON Parsing, Shell scripting)- Recommended additional features on top of ASR which received appreciation from customer.- Received appreciation from customers for improving the efficiency of the code by emphasizing on code‑reusability.- Managed the ASR project as a single point of contact.- Developed a PoC for Emotion Detection and Sentiment Detection from text documents using transferlearning on BERT LM (Dataset used: IEMOCAP, MELD)- Delivered document similarity using doc2vec, word mover distance and LSI method with a detailedreport on all the approaches used.- Implemented Entity extraction (NLP technique) from the given search query and mapped SQL datausing Neo4j with graph visualization.- Enhanced the Recommendation System by reducing the space complexity and successfullyperformed JMeter load test.- Completed a capstone project on image classification as a trainee Show less

  • Senior Data Scientist

    Jun 2021 - Aug 2021

  • Data Scientist

    Sept 2018 - May 2021
• 

  NielsenIQ

  Sept 2022 - now

  Sr. Expert Innovation and Automation

  - Worked on Semantic Search (Retrieval, Encoder and Reranker Methods) to extractbest matching product out of millions of products available.- Created annotation pipeline and training pipeline for Object Detection model toextract Promotions/Description/Products from a flyer.- Trained NER model to extract Brand, Quantity, UoM and Prices from Descriptiontext above.- Saved around 14+ FTEs (estimated) by building multiple Automation involvingETLs.- Deployed string-matching using TF-IDF on character level using Streamlit.- Multilabel text classification using FastText on promotions product classes. Show less