• with state-of-the-art transfer learning models (Feature Extraction and Fine-tuning)
• with customized text preprocessing logic using computational linguistic techniques wherever it helps!
• which are deployed using Python CLI apps, FastAPI REST APIs or Streamlit UIs in Kubernetes

• which is reusable, replicable and runnable in docker containers
• which is modularized and packaged (as from some_internal_package import what_you_need)
• committed to GitHub for co-development and issue-resolution
• with docstrings and pytests, subjected to Pull Requests when multiple developers are involved

	Extensively Used	Working Knowledge
Tools
Python Libraries

• WSL for local development, and linux machines for GPU-powered, dockerized applications development
• predominantly PyCharm (Professional) for remote development but use opensource VS code for local development
• Jupyter Notebook to learn coding concepts
• draw.io, mermaid and markdown for flowcharts and documentation purposes

• OneNote for taking lots of notes from emails, meetings and websites and
• Slack for communicating, weekly updates and jotting down reminder messages to self
• Microsoft PPT for conveying data stories/insights to non-developer team mates or superiors

• A People Manager
  • directly managing the delivery of Social Media Analytics projects of 8+ members in my stint at LatentView Analytics
• A Technical Mentor/Trainer
  • enhancing the NLP/Python expertise of fellow team members or reportees

• Built a reusable Sequence Classification ML Pipeline which converts customer comments into Aspects and Sentiment
• Highlights of the Pipeline:
  • Spark+Spacy Preprocessing
  • Transfer Learning + Clustering aided Annotation
    • Less annotation for Training (compared to traditional ML) by intelligent use of DL+ML models
  • Dockerized Environment for Model Training and Inference
  • Fine-tuned Transformer models
• Look here for more details

• Annonymized PII in text data that resulted in less restricted use of the data
  • by building a Named Entity Recognition (NER) system that can detect PII
• Highlights of the Pipeline:
  • Bootstrapped the training data using Spacy rules (thus easing the annotation process)
  • Spacy's Roberta Base Transformer model allowed for no truncation of sentence max length
  • Inference REST API (via an asynchronous FastAPI deployment using K8s) that can be plugged into multiple applications
• Look here for more details

• Goal: To create "digital threads" for connecting automotive data sources

  • which has technician comments about issues before the launch of a vehicle,
  • by assigning semantically matching common part categories to every issue in both data sources

• Built a pipeline that ensembles results of 3 pairs of Retriever-Reader models wherein

  • the Retriever narrows down the search space and
  • the Reader zeroes in on the right results

• Look here for more details

• Built reusable Text Clustering pipelines
  • for deriving actionable insights from unlabeled text corpus
• Highlights of the Pipeline:
  • The clustering pipeline provided options for both Traditional Topic Modeling and DL-Embedding based Hard Clustering
  • Incorporated the models into an easy-to-use Streamlit UI deployed via K8s
  • The codebase was built on top of the main open source libraries
    • PyTorch (Transformers, Sentence Transformers) and Sklearn
• Look here for more details

How did I start my career?

• Back in July 2010, I had started out providing customized Market Research (MR) in my first 4 years of my career.
  • Simply put, it was a no-code work
    • involving cold-calling, speaking to experts and reading a lot of secondary research material
    • to write actionable procurement intelligence reports .
  • This first job, right after my engineering undergraduation,
    • had taught me the importance of tough-to-learn soft skills
    • especially in communication be it written, one-on-one, cold-calling, team presentations and many more.

When did I transition to NLP?

• Since 2014, I have been in the field of Data Science, and the romance has not died down yet :).
• Largely because of the interesting NLP opportunities that landed my way.
  
• I had primarily worked on Social Media Analytics at LatentView from 2014 to 2018 where
  
  • I had aided my F100 tech major client to effectively use social media insights in their marketing decisions
• Since May 2018, as a Data Scientist at Ford,
  • my technical learnings in ML/DL and NLP have been on an upward trend!

What are my mottos?

Striving to follow the below mottos for professional betterment:

• To keep upskilling my technical knowledge
  • Firmly believe there are Miles to go before I sleep
    
• To bring the best collaborative, transparent and importantly humble self in my interactions with colleagues/friends,
  • This is so that trust is enabled, long-term partnerships are forged and great results are achieved
    
• To stand on the shoulders of the giants of open source
  • In other words, be applied practitioner first, and not try to reinvent the wheel unless it has some learning/business benefit

My Small World

• I am here working happily in the Data Science field largely because of the sacrifice & guidance of my wife .

  • She guided my transition from Market Research to Data Science. She is a fellow analytics professional too
    
  • She is on a break to take care of our possibly autistic todler son.
    
  • I am cognizant of this privilege that I am enjoying (me being able to work when she couldn't).
    
  • It has been particularly exacerbated by covid situation and personal losses
    

• Speaking of my son
  

  • He is the apple of my eye
    
  • He seems to have exemplary memory, well beyond his age! (possibly biased opinion 🙂)
  • He grasps abstract things like shapes, numbers, letters, and words faster
  • He could be in some autism spectrum (slower learning in social skills compared to kids of his age)
    - With my wife's leadership we diagnosed it early and
    - Hopefully we are acting on it early before it becoming too noticeable

My Interests

• For last 2 years, I have spent (okay, wasted!) a lot of time on many must-watch TV series. Some iconic I must say.
  
  • My favorite genres: Sci-Fi, Comics, Legal/Medical thrillers and anything out of this world
    
• My favorites among novels include many mythology fictional writings
  
• An ardent tea lover!

• Here is my résumé in pdf