Sukhman Singh

Experienced Data Scientist/ ML Engineer having 4+ years of experience in successfully building production ready NLP models, microservices and hands on experience on Machine Learning, Statistical Modeling, Predictive Analysis, Data preprocessing, AI/ML Dashboards, Data Visualization, Prompt Engineering.

• Built text classification model to tag the behavioral tactics automatically whenever a behavioral scientist adds a new copy.
• Utilized pre-trained text generator models like ChatGPT and Davinci to generate, rewrite, and summarize emails and SMS outreaches with efficient and engaging prompt engineering.
• Developing quick prototypes using Gradio to showcase ML model abilities to internal stakeholders. Deployed all these apps on Django with Databricks job scheduler.
• Worked on deploying apps as an Azure web application with Microsoft SSO authentication.
• Built generalizable data pipelines used across multiple clients with differing data models. These pipelines generate the core data science tables used in both ad-hoc analysis and model development.
• Containerized many different applications to make it deployable and useable across different platforms using Dockers.
• Converted our workflow pipelines into installable python packages, making it easy use across multiple teams.
• Built error analysis dashboard package to display various evaluation matrices like classification report, custom confusion matrix, Category distribution pie charts and deep keyword analysis.
• Built keyword analysis algorithms with the help of topic modeling methodology using TF-IDF vectorizers and count vectorizers.  
• Working on Text classification and Sentiment Analysis projects to build production ready models to classify customers problems, intents and their sentiments in the SMSs and emails.  
• Improved state-of-the-art BERT model for sentiment analysis with Data Centric approaches, by improving annotation and preprocessing.
• Worked with spacy pipelines for preprocessing and building masking techniques to mask PII data from customer SMS and emails. 
• Utilized auto documentation in python using libraries like pdoc3 and sphinx, to generate code documentation from docstrings. 
• Rapidly updating and writing documentations for the work done, to make it understandable for the current and new team members.  
• Built natural language processing pipeline using NLTK, Regex, pandas in order to clean text data(training-data) for multi-labeled Text Classification. 
• Built PDF-TO-OCR pipeline using pdf to image converter and pytesseract to convert images to text files with custom bounding boxes to extract tables from it.
• Multi GPU batch processing for fast training with bigger batch size in TensorFlow and Pytorch.
• Worked with Kubernetes and docker for containerization to deploy deep-learning models locally and in cloud as a FLASK REST API.
• Built algorithms like Facial-Recognition, Anomaly detection, Generative Adversarial Networks (GANs), Autoencoders, Variational Autoencoders.
• Worked with processing live streaming data from PUB/SUB systems like MongoDB, PostgreSQL, writing T SQL queries to create functions and triggers.

Currently pursuing MSc in Data Science(Remote) from Liverpool John Moores University, UK