Prateep Sengupta

• Working as the Lead Data Scientist for the project 'Watson Live Captioning (WLC)', intended for Sinclair Broadcasting Group.
• Working mainly in the domain(s) of Speaker Change Detection and Punctuation Restoration.
• Created a speaker change detection evaluation automation script and integrated it as a functionality for the existing evaluation pipeline for WLC as a whole.
• Worked with speechbrain, an open source speech framework, and used their speaker recognition system as the base of our next gen speaker change detection system.
• Led the model training part of speechbrain as well as data preparation for training and created model evaluation pipeline.
• Was part of the model deployment on AWS Sagemaker as well, which makes our Data Science+Engineering team to arguably, one of the first teams to have successfully deploy speechbrain's
  custom trained model in production.
• Apart from that, I have also worked on the development of a next generation punctuation restoration model using LLMs such as RoBERTa and Distill-BERT.
• Worked on two different algorithm to estimate automatic F1 Evaluation for SCD as well as Punctuation Restoration - one is based on sentence vectorization & sentence similarity, and other (the latest one) based on word level match using word timestamps.
• Currently, I work on and oversee the progress of both the SCD as well as the Punctuation Restoration Model training and evaluation.
• Also, currently planning on having expert level certifications on Deep NLP and Generative AI.

Joined the department of BDGS in their OSS Analytics team and worked till early Oct, 2021. I worked there mostly as a Python SME as well as in GCP related development for Ericsson Product RTPM (client: MBNL UK), an AI based alarm and monitoring system based on Google Cloud.

1. Worked as a Python/ML Expert for the development of the Ericsson product VILA (Visually Impaired Life Assistant). In this project, I have worked on the following-
Unknown Face Clustering,
Face Recognition / Facial Analytics,
Face Image upload and aggregation pipeline using Kafka/Spark
Menu card Reading using OCR
SIFT, SURF and ORB based pattern recognition
Related REST Services using Django-REST Framework

2. Worked as a core team member (Operations and L2 Support) of the Kolkata team for the Ericsson product EEA (Ericsson Expert Analytics).

3. Worked as a Python SME for the development of a British Telecom specific customization on the Ericsson DevOps product Rosetta (Django)

4. Worked on Ericsson product CSD (Cognitive Support Desk), to develop a machine learning based solution as well as REST backend for clustering the customer emails for efficient grouping.

5. Worked on a hackathon project to develop a deep learning based solution to predict the estimated hours to complete multiple tasks, given LLD level data. The solution was selected till the second screening of the hackathon.

• Research & Development for a POC for the
  advancement of the existing Email-Bot framework.
  (Zoe-Email). The stacks used are Flask (Python) and
  Google Cloud's AutoML apis.
• Research and Development for a chatbot POC
  from scratch using RASA stack ( Rasa NLU & Rasa
  Core)
• Maintenance, Training and Testing chatbots built
  on in-house platforms (Zoe, e2m) for various
  clients.
• Was involved in setup, training and maintainance
  of chatbots for the following clients:
  Customer Bot for Visa
  Les 3 Fontaines Mall ( Hammerson UK)
  Event Bots for Reed Exhibitions ( Vision Expo West,
  G2E 2018, Chicago Comic Conference, to name a few)

1. Web Data Mining for PatientMD's Doctor App,
using Scrapy & Selenium.
2. A concurrent streaming application of stock
details from Google Finance using Akka
Framework.
3. Research / Development of a sentiment
analyzer using Python libraries such as TextBlob,
Stanford CoreNLP, NLTK, Spacy etc.
4. Research / Development of a Speech
Recognition Module which is to be integrated in
our healthcare chatbot as a speech-to-text
service (as a Flask REST service)
5. Some debugging and occasional development
of the core backend functionalities. Have
developed a first draft of their shopping cart API
for genomics products. The framework used is Play
(Scala).

1. Project Management
2. Client Management / Client Handling
3. Technology Research
4. Email Marketing (Zoho, MailChimp, Agile CRM)
5. Web Data Mining & Web Content Mining
6. Excel VBA Development

Following were my responsibilities-
1. Web Scraping / Web Crawling
2. Macro development using Excel VBA
3. Network / Server troubleshooting
4. Email Marketing
5. Secondary Research

Worked as a Junior Project Assistant in IIT KGP in
the project entitled "Automatic Speaker
Recognition on VoIP" sponsored by Vodafone
Essar Ltd. Have built a first draft of a speaker
recognition system over Skype as a VOIP sysyem.
Technology Stack: bash scripting, Perl, C