Dec 2021 – May 2024 · 2 yrs 5 mos · Bangalore Urban, Karnataka, India · Hybrid

• Working on upgrading Customer Service Requests tool with machine learning, NLP and
• statistical concepts
• Domain Specific SBert Fine Tuning
• The goal is to improve the overall accuracy of clustering and classification models by fine
• tuning Sentence-BERT (SBERT) using Masked Language Modeling (MLM) technique to
• generate high-quality sentence embeddings. Implemented the pretraining pipeline for SBERT using MLM, leveraging the PyTorch framework and Oracle’s Service Requests tickets
• Service Requests Topic Prediction
• The support team had a requirement to monitor the daily trends of new service requests.
• To address this, machine learning models were developed to classify service requests into
• high priority topics or clusters, as provided by users. To make the models easily accessible and scalable, an API endpoint was developed using Flask-RESTful
• Service Requests Topic Clustering
• The goal of this project is to enhance customer service by effectively clustering and
• analyzing service requests. By leveraging Agglomerative Hierarchical clustering, the tool
• categorizes incoming service requests into distinct clusters. This enables support teams
• to prioritize and address different types of customer issues, leading to targeted support
• and product improvement plans.

Nov 2019 – Dec 2021 · 2 yrs 1 mo · Gurgaon, India

• Working on building Data Science tools and democratizing it within the organization to help answer clients business queries and problems.
• Enabling different analytical modules based on statistical measures to help analysts discover trends and derive knowledge for business use cases.
• Research Work
• Identifying Innovation/Growth Signals in Company Portfolio description and earning call transcripts. Working on fine tuning FinBERT (BERT model pre-trained on financial communication text) using earning calls transcripts dataset. We have used Keyword Tagging capability ( an in house product) to create innovation based labelled dataset. Aim is to create a FinBERT based classifier that can predict and label documents as Innovative or Non Innovative entities. This prototype can help with decision making related to merger & acquisitions and scoring of a company's Innovation proficiency.
• Keyword Tagging is a python based tool, created to tag/label excel documents against query-based mechanisms. A Hybrid algorithm was developed based on Boolean, Proximity and Wildcard searches mechanism, similar to one which are used in Google or Microsoft search engines. It works well with overlapping data where clustering fails to segregate data properly.
• Case Work
• Predicting Merger & Acquisition Targets: Built a funding/revenue based statistical measure along with a ML based classification model to recommend clients with AI based companies for either investment or acquisition.
• Finding Adjacent Growth Opportunities: Used patents dataset to create a statistical measure which computes the relative difference of a companies Patents. For relatively distant patents, used Topic Modelling (LDA) on patent description to identify new growth areas for different companies.

Dec 2017 – Nov 2019 · 1 yr 11 mos · Gurgaon, India

• In Fidelity International, I had been enrolled in complete end to end Product Development.
• I had worked upon projects consisting of Big Data Technologies such as Python, PySpark, Pandas, Numpy, Oracle, MongoDB etc.
• Projects are as follow:
• Investsense - Analytical Engine which analyse customer's data and Identify their needs for Investment, recommends investment products, analyze current portfolio performance and suggest better products
• GDPR - Analytical engine to identify if Customers Data residing with in the organization is adhering with GDPR policies and rules
• Crypto-League - A web-platform to invest in Crypto currencies. It was a game designed for Fidelity Employees to learn about Crypto investment and technology. My role was to build Data Models using Python and MongoDB and back end services.
• Reconciliation - Built an Automated Python Framework to compare Millions of Data from different Data Sources and reporting mismatched data to DB team.

Aug 2016 – Dec 2017 · 1 yr 4 mos · Gurgaon, India

• Built Real Time Adhoc Reporting and BI solutions .
• Preparing Data Models
• Analyzing data using Python Rest API's
• Reporting using BIRT visualization tool
• Asset Management System
• Developed Asset Management System for G4S in which Employee data along with the
• assigned machine(Computer) data is automatically submitted to the database using their Email.
• Developed Back end API, in which commands were written to collect data from any employees
• machine. Data collected if further processed and cleaned in Python and finally submitted to the
• database.
• Technology Being Used : Python, Mysql, GIT, Google Cloud Platform, Nosql

Jun 2016 – Aug 2016 · 2 mos · Sector-48 Gurgaon

• Developed a Digital Magazine application for retail vendors wherein the owner can offer their products and services and customer can make reviews for the same. It was an application designed to be used even without Internet connectivity. The technologies used to build this application :– Firebase, Angular Js and Ionic Framework, Angular Material.

Oct 2015 – Nov 2015 · 1 mo · Gurgaon, India

• Worked under the guidance of CTO Rakesh Verma at Valion Realty, Gurgaon.
• Developed ETL based python scripts to manage database.

Jun 2015 – Jul 2015 · 1 mo · New Delhi Area, India

• Developed Staff Selection Commission website modules.
• Technology used : Google Web Toolkit, JAVA, CSS 3, HTML, Photoshop

Jul 2014 – Aug 2014 · 1 mo · Indore Area, India

• Developed Asset Management Project
• Technology used: PHP, HTML, JSON, MySQL.