Sidharth's Profile

About

Data science professional with 7+ years of experience. I am currently in financial sector driving sales and marketing analytics. I have supervised a team on building products with focus on alternative datasets using latest technologies. Also I have helped teams to pivot from legacy systems though AI and automations. Apart from this I have experience in US healthcare domain as well. I have proven academic records and demonstrated analytical capabilities on competitive platforms. I belong to Muzaffarpur, Bihar in India

Personal Details

Birthday: xx - xx - 1994
Website: sidharthiimc.github.io
Martial Status: Family Man
Current Location: Hyderabad, India

Age: 30
Degree: Master
Email: sidharth.kr@hotmail.com
Freelance/For Hire: Available

I have experienced living in Hyderabad, Bangalore, Delhi, Noida, Kolkata as well. Watching general studies lectures on YouTube is my favourite free time activity. I like travelling a lot. I have trekked up to 5000m above sea level. I also like to play cricket and badminton. Tea is my favourite beverage.

Facts

I am into Data Science from mid 2016. I have not only learnt but also guided several folks. Many of them have reached even better position than I am today :) . Key highlights in the professional experience -

Years of Experience

Projects delivered

Mentored Students and Colleagues

Followers and growing

Skills

Most of my skill come from my Post Graduation in Business Analytics, Corporate Work experience and Solving challenges on Kaggle. I still visit kaggle to checkout latest development in Data Science and Machine Learning.

Python 99%

R 75%

SQL 90%

Flask - Web App and API 65%

Office (Excel, Powerpoint) 80%

Data Science 90%

Machine Learning - Supervised95%

Machine Learning - Unsupervised95%

Deep Learning 70%

Image and Video Processing 65%

Natural Langugae Processing(NLP) 90%

Generative AI70%

Resume

Summary

Sidharth Kumar

Data Science Professional with 7+ years of industry experience. Currently working in Financial sector with focus on sales and marketing. Also has experience in Market Research (alternative data) and US healthcare domain. Proven academic records and demonstrated analytical capabilities on competitive platform.

Based in Hyderabad, India
(123) 456-7891
sidharth.kr@hotmail.com

Education

Post Graduate Diploma in Business Analytics ( PGDBA )

2016 - 2018

IIM Calcutta - ISI Kolkata - IIT Kharagpur

Specialization: Data Science, Machine Learning, Business Analytics

Key Courses: Regression, Algorithms Design, Machine learning, Information Retrieval, Healthcare and Product Analytics, Stochastic processes, Statistical Data structures, Inference, Computing for data sciences, Database Systems, Categorical data analysis, Data Mining,Project management, Econometrics, FRM, Computational Finance

CGPA: 8.74

Bachelor of Technology (B.Tech)

2011 - 2015

National Institute of Technology (NIT), Patna

Specialization: Mechanical Engineering

CGPA: 8.80

Class XI- XII

2009 - 2011

Central Board of Secondary Education (C.B.S.E)

Score: 91%

Class X

2008 - 2009

Central Board of Secondary Education (C.B.S.E)

Score: 90.5%

Professional Experience

Data Scientist

Jan 2024 - Present

Franklin Templeton, Hyderabad, India

Development of behavioral personas for US financial advisors based on marketing signals
Tech stacks exposure - Python, Snowflake, DataBricks, Power BI

Senior Data Scientist

July 2019 - Dec 2023

GlobalData Plc, Hyderabad, India

Development of AI based applications using GPT and fine tuning LLMS
Development and implemented several predictive anayltics solutions
Working on NLP technology stack for alternative datasets like Twitter, Jobs, Company Filings, News, Reddit
Contributing to business development by providing innovative solutions
Developed several UI and Dashboards for sales and marketing
Tech stacks exposure - Python, Transformers, Google-Big Query, Data Studio, Plotly Dash and Flask framework

Data Scientist

April 2018 - June 2019

UnitedHelath Group, Noida, India

Provided solution to US-healthcare in Medicare and Marketing business.
Deleloped algorithms for NPS star ratings, CAHPS, Retention Programs
Analysed Large dataset to find key insights
Tech stacks exposure: Python, R, Sci-kit learn ML models, Bigdata- Hive

Internship - Business Analytics

October 2017 - April 2018

UnitedHelath Group, Noida, India

Provided solution to US-healthcare in Provider and Digital business.
Deleloped algorithms for digital revolution
Automated data extraction , transformation and loading Process
Tech stacks exposure: Python, Bigdata- Hive and PySpark

Software Engineering Associate

January 2016 - May 2016

Accenture, Bangalore & Hyderabad, India

Worked for US-based software client to make web applications accessibility compliant
Prepared test cases & scripts, ran test scripts, and logged bugs

Download PDF version of my resume from here: Resume

Projects

I have worked on variety of project throughout my career. I have projects on core data science to data analytics and even data engineering. Majority of my projects are on Tabular data and Natural Language Processing. List of project and roles at various organisations and institutes -

All
Franklin Tempelton
GlobalData
UnitedHealth Group
Academics
Kaggle

OPTUM - Data Science Challenge

1st Runner up world-wide (1st in India) for wining data science challenge among employees

Built a PCP recommendation model that can reduce cost by choosing right PCP for members

Final Model was built using 10 variables with lots of advanced feature engineering technique

Achieved an AUC of 86.66 on public leader board using a single model with tuned parameters

Financial Advisors Segmentation

Developing segments for US finaical advisors based on their interation with various marketing signals

Nine robust clusters were identified using PySpark Kmeans

Cluster labels were generated using deep dive interpretation of each cluster by projecing various features and metrics

Power BI dashboard was built for sales executive usages and understanding of various segments

Multi-Task Private LLM

Developing consumer grade generative AI capabilities by fine-tuning LLM for various operational task like summarization, tagging, data extraction.

Ensuring private usage of LLMs by for data privacy and safety

Trend Pusle

Built a daily newsletter and dashboard for real time trend identification from social data along with alert mechanism.

Built a trend explanier using GPT and automated timeline peak identification

Achieved full automation of the process using mail chimp, elastic search, g-sheets, data studio, email APIs

Sense Hub

Developed a Question Answers and report writing bot using GPT automated agents

Built a semantic search engine for retrieval and used GPT for Q/A. Created multi-layered chat & semantic engine using agent bots to write quality reports using own dataset.

Added organisations existing analysis frameworks in the AI search to use it in Q/A response.

Patents Company Name Normalisation

Created an algorithm to map company name variants with an existing company naming system

Utilised text cleaning, patent features like family & inventor details and company features like headquarter, revenue and several user defined metrics for developing the matching algorithm

Algorithm was able to cover 95% of all the patent applicants. New variant database can be utilized in various other data sources like news, deals and social for faster information processing

Retention Driver Analysis

Worked on finding key drivers influencing member retention during annual enrolment period

Built logistic, random forest and gradient boosting model for identifying key driving factors

Utilising concept of shapley value and partial dependent plots for quantifying the drivers impact

Consumer Review Insights

Created an automated cluster and topic generation for driving insights from drug reviews data.

Used Hugging Face Transformer for vectorising and density based hierarchical clustering.

For topic generation used spacy with zero-shot similarity metric

Power Point Automation

Using Python-pptx libarary, automated 5000+ ppts.

This helped in JIT PPT geneartion which helped in cost and space reduction

Stock Market Movement Prediction using Social Sentiments

Modelled the social sentiments on stock price to predict the direction for the next trading session

Extracted features using Information Retrieval based techniques from the top 25 daily news data

Achieved an accuracy of 59.52% on test set using SVM and simulated a buy/sell trading strategy

PR Reports, POCs, Automations

Created a report on Vegan based on social data using text and keyword search on the corpus

Created a report & interactive dashboard on beers to show case capabilities of utilising blogs data

Automation of several legacy one pager reports using python to save manual time and efforts

CAHPS Project

Worked for enhancing NPS & Medicare Plan Star ratings by utilising the member survey data

Analysed contract & plans star rating and likelihood to reach 4 star rating using logistic classifier

Predicted members who are likely to provide low scores in survey using random forest model

Retrieval based Chat Bots

Developed an algorithm for retrieval based chat bot on public available Ubuntu Dialogue Corpus

Applied methods that include Cosine Model-TF-IDF using n-grams, Dual Encoder LSTM (RNN)

Achieved a recall measure of 77% using the Cosine model and a recall of 88% using LSTM(RNN)

Analytical Dashboards

Verdict Media websites traffic monitoring for sales team using Google BigQuery and Data Studio

Dashboard using social indicators to analyse the effect of Covid-19 on companies and sectors

Competitive Intelligence Dashboard for company performance benchmarking using Plotly dash

Created a UI for Analytics Hub dashboard for analysis & metric generation from alternative data

Member Tenure Prediction

Worked for personalising prospect acquisition, member engagement and retention initiatives

Worked on random forest survival model for predicting member churn and its tenure with UHC

Performed Customer Life Time Value analysis using the predicted tenure for marketing strategy

Disruptor Innovation Identification

Created an ML model to classify an article into innovation or not.

By fine tuning BERT embedding, an accuracy of 93% was achieved.

Also built a second stage model to summarise the article by finding benefits and challenges in concerned innovation

Robust Regression using Campbell's Method

Assigned proper weights to observations of many public datasets by iterative Campbell's method

Found the outliers using the weights and implemented weighted regression for prediction in R

Analysed the properties of the Campbell’s method of estimation using Monte Carlo simulations

Internship - Provider Digital Analytics

Worked for Provider side of the business to improve experience and reduce cost by digitisation

Project involved working on Big Data platforms like Hive & Spark for data collection & analysis

Created a complete automation of process from data ingestion to data transformation and update

Received Bravos 2 times from the Product Owner & the Manager for the collaborative work done

Received PPO to serve as full time data scientist & graded excellent in the Internship evaluation

Influencer Recognition

Created a filter identifying individuals using the profile images from twitter for identifying the social media influencers.

Model was built using Convolutional Neural Network (CNN) with an accuracy of 95% and reducing manual effort by more than 90%

Sentiment Model

Built an ML model for measuring the sentiments of tweets using the internal and public data.

Gathered 3 million training data from various open sources like universities & data platforms.

Utilised the word2vec approach and deep neural net for training the sentiments

Sberbank Russian Housing Market
Price Prediction of housings in Russian market, used extensive data cleaning and applied an ensemble learning approach for modelling the price

Booz Allen Hamilton’s Data Science Bowl
Lung cancer prediction from the medical scans, used Xgboost and Neural Network for the purpose of prediction to save life and to reduce cost

Quora Question Pairs
Modelling Semantic Similarity of question asked on Quora, used LSTM and Xgboost along with feature engineering to find similar questions in the test set

Sea Lion Population Count by NOAA Fisheries
Used CNN for counting sea lions in multiple categories in drone images

Fraud transaction detection identification by IEEE
Analysed data and identified fraud transaction using predictive modelling

Z-estimate for Zilliow
Predicted US home price by modelling varibale to upgrade the z-estimate

Kaggle kernel only competition Instant Gratification
Worked on artificial dataset to decode the pattern using advace algorithms