AI Engineer and Data Scientist with 7+ years of experience specializing in the financial technology domain. I focus on architecting production-grade machine learning pipelines, deploying Generative AI applications, and building scalable data infrastructure that drives business growth and operational efficiency.
Core Technical Stack:
Machine Learning & AI: Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), Agentic Workflows, Natural Language Processing (NLP), and Predictive Modeling.
Engineering & Frameworks: Python (NumPy, Pandas, Scikit-learn), PyTorch, SQL, LangGraph, and RASA.
Professional Experience & Impact:
Throughout my career, I have bridged the gap between raw data and automated strategy. In the Fintech sector, I architected and deployed an NLP-driven chatbot using the RASA framework, drastically reducing operational overhead by automating high-volume customer service inquiries.
Currently, I engineer automated data preparation and manipulation pipelines to extract and structure Andhra Pradesh Commercial Taxes data, ensuring critical, high-stakes decision-making is backed by accurate analytics.
Open Source & Community:
Beyond my day-to-day role, I am an active contributor to core open-source AI frameworks, such as LangGraph, focusing on resolving backend infrastructure issues, including asynchronous memory leaks. I regularly document these architectural deep dives, LLM deployment strategies, and system design learnings on my technical Medium blog.
I am currently targeting remote Senior AI Engineering and Data Science roles where I can leverage my domain expertise and modern AI infrastructure skills to build innovative, highly scalable products.
Experience
Data Analyst
Andhra Pradesh Centre for Financial Systems and Services (APCFSS)
3 yrs 5 mos · Vijayawada, Andhra Pradesh, India
Technical Writing Intern at Medium, Showwcase, and Hashnode
Prototyped a conversational NLP chatbot using the RASA framework to dynamically handle regulatory compliance FAQs for financial institutions. Built a custom intent recognition engine utilizing entity co-occurrence strategies, improving model confidence levels from 40% to 80% and achieving an 85% match rate against human baselines. Developed automated, high-throughput web scraping infrastructure using Python and Pandas, increasing data extraction efficiency by 90% to feed downstream ML training pipelines.
Python Teacher
ReDI School of Digital Integration
3 mos
Python Teaching
Virtual Volunteering
Data Scientist
OPERANKA ASSOCIATES
3 yrs · Greater Paris Metropolitan Region
I’m coming back from a sabbatical & virtual volunteering year at ReDI School of Digital Integration in India as a Python tutor where I did data research projects using Pandas, SQL, Numpy, and Power BI. Before that, I worked in Paris at Operanka Associates, Paris, France
Project: Operanka Associates is a Paris-based FinTech startup delivering financial data analysis & artificial intelligence services to banks like HSBC, BNP Paribas & Société Générale. I worked in a data team of 6 creating, cleaning, and preparing client datasets using Python, SQL, Pandas, Numpy, Tableau, and Power Bi.
Analyzed a data set of 50k+ customer questions to improve customer satisfaction metrics and created various dashboards using Power BI. Developed solutions by web scraping using Python to extract data e.g. the texts, links of websites, and relevant documents of data protection regulations from the internet. This reduces the effort for manual data extraction by 90%.
Automated the process of information extraction of EU data protection regulation text by 2 to 3 hours instead of manual google search of files.
Developed a chatbot using NLP. Improved the confidence level of intents and entities from 40% to 80% by creating co-occurrences of certain entities which involved handling lists, and string data types.
Tech used: Python | SQL | Pandas | Numpy | Power Bi | R | NLP | Data Analysis | NLP Web Scraping | Jupyter Notebook | Linux | Data Cleaning | Data Preparation.
Data Analyst
OPERANKA ASSOCIATES
3 yrs
Engineered an NLP-driven recommendation engine to resolve user connectivity bottlenecks, matching 1,000+ users based on semantic interest similarity and driving a 50% increase in platform interactions. Architected automated data pipelines integrating multiple APIs to aggregate and process real-time scheduling data from 500+ transport networks, optimizing cross-modal navigation features.
IT and Data Science Recruiter
Talent Is Everywhere (TIE)
2 mos · Greater Lille Metropolitan Area
Worked on connecting the right IT and Data Science candidate to their right job description.
Data Analyst
Wegravit SAS
6 mos · Greater Lille Metropolitan Area
Project: Wegravit a Lille-based startup that created a social media engine without leveraging advertisements and not utilizing user data to make profits.
I worked in a Data Analyst team of 3 developing a filtering algorithm to recommend and match users to other users according to the specific keywords (e.g. similar hobbies, same University, work domain, etc.) on the social media platform using R, Python, SQL, and recommendation techniques. This helped users find the right connections to socialize.
Designed a transportation and navigation application (as part of a competition by the SNCF train company) to assemble all kinds of transport (accessing real-time traffic data of metro, tram, bus, and cycle of the city of Lille) within one single application along with other added features using tools like R, Python, JavaScript, and PHP.
Tech used: R | SQL | Python | Recommendation systems | Javascript | PHP | Data Analysis | Pandas | ggmap | ggplot2.
Data Science Intern
6 mos
Assistant Professor
GMIT | Gargi Memorial Institute of Technology
1 yr · Kolkata, West Bengal, India
Conducted courses Circuit Theory, Signals and Systems, Digital Signal Processing and laboratory work on MATLAB software. In addition I was also involved in organizing laboratory experiments on hardware for Analog Electronics.
Data Analyst and Scientist
APCFSS
Jan 2023-Current · Vijayawada
Engineered automated data extraction and processing pipelines using Python and NumPy, optimizing taxpayer dataset handling by 50% across 100,000 high-dimensional records; implemented Levenshtein-distance fuzzy matching for address resolution across 5+ government departments; collaborated with cross-functional technical teams to reduce project delivery latency by 3 to 4 weeks.
Data Analyst & Scientist
Operanka Associates
Jun 2018-May 2021
Prototyped a RASA conversational NLP chatbot for regulatory compliance FAQs; built an intent recognition engine that improved model confidence from 40% to 80% and achieved an 85% match rate against human baselines; developed Python and Pandas web-scraping infrastructure that increased data extraction efficiency by 90%.
Data Analyst
Wegravit
Jul 2017-Dec 2017
Engineered an NLP-driven recommendation engine matching more than 1,000 users by semantic interest similarity and increasing platform interactions by 50%; built automated API-integrated data pipelines processing real-time scheduling data from 500+ transport networks.
Education
Johns Hopkins Whiting School of Engineering
Certification, Applied Generative AI
Oct 2025 - Mar 2026
IÉSEG School of Management
Master of Science (MSc) in Big Data Analytics, Big Data Analytics for Business
2016 - 2017
Predictive Analytics in Fund-raising: - A predictive model was made to predict which donors will respond to the reactivation campaign. The models' Logistic Regression, Decision Tree and Regression Tree were made to find out the best predictive model using R.
Inventory Optimization:- We estimated the amount of sales using a forecasting time series model for the upcoming months for Leroy Merlin and represented the results in an interactive dashboard by Shiny R.
Analysis and Visualization of Airplane Safety data by Tableau:- Investigated the safety of airplanes for a newspaper article. To support this objective, the data visualization tool Tableau was used to explore the data and discover interesting facts thus creating relevant visualizations. Moreover, the analyses of the findings and visualizations were supported by the concrete significance and proper conclusions.
NIIT University
Research Scholar
Jan 2013 - Sep 2015
Project: NIIT University, private engineering University in India which focuses on research and development-based learning environments. Worked as a research scholar with 4 other research scientists on the topics of speech signal processing and communication systems. I analyzed speech accents of non-native English speakers from northern India and eastern India on certain English words and then suggested remedial services to aid first-hand accurate language acquisition using MATLAB and Wave-surfer speech analysis simulator (Open source tool for sound visualization and manipulation created by KTH, Sweden).
Tech used: MATLAB | Wavesurfer | Speech Signal Processing | Voice accent recognition
Heritage Institute of Technology
M.Tech, ECE
2010 - 2012
Maulana Abul Kalam Azad University of Technology, West Bengal formerly WBUT
Bachelor of Technology - BTech, Electrical
2006 - 2010
Forage
KPMG Data Analytics Virtual Experience Program
May 2022
Language Pantheon
German Language Course, German Language
Aug 2021
Youtube
Data Science and Analytics
May 2021
IÉSEG School of Management
Master of Science, Data Processing and Data Processing Technology/Technician
[object Object]
MAKAUT (WB)
Bachelor of Technology, Electrical
[object Object]
IESEG School of Management
Master of Science, Big Data Analytics for Business
Aug 2016-Aug 2018
MAKAUT University
Bachelor of Technology, Electronics and Communication Engineering
Aug 2006-Aug 2010
Skills
Pandas (Software)PythonNumPyData ScienceData AnalyticsLarge Language Models (LLMs)Retrieval-Augmented Generation (RAG)Agentic WorkflowsNatural Language Processing (NLP)Predictive ModelingPyTorchRASAPandasScikit-LearnSQLAPI IntegrationVector DatabasesLocal Model DeploymentData Pipeline Architecture
Projects
Business Reporting
The goal of the assignment was to create a business report for a small translation agency analyzing the various attributes about the project and gain insights from it using SQL. The project involved cleaning the data, analyzing it and finally representing the data results in a dashboard by Tableau.
Predictive Analytics in Fund-raising (2016)
A predictive model was made to predict which donors will respond to the reactivation campaign. The models Logistic regression, Decision Tree and Regression Tree were made to find out the best predictive model for the problem using R.
Advanced Spreadsheet Analysis (2016)
VBA codes and macro were used to display the performance of sales stores on the excel dashboard.
Credit Scoring
An application credit scorecard was built where the raw data represents a sample of accepted and rejected applicants, as well as a description of the data. There are steps about how the model is
being made by the help of SAS Enterprise Miner and the reason behind following the steps in the proper manner based on data science and evaluation of the choices being performed using statistical measures. The steps and choices were explained thoroughly that have been made in building the application scorecard, as well as detailed discussion of the scorecard and its performance evaluation was done.
Inventory Optimization
Estimated the amount of sales using a forecasting time series model for the upcoming months for third largest DIY retailer using R and represented the results in an interactive dashboard by Shiny R.
Data Visualization by R on the WHO dataset of Healthy life expectancy
The project involved representing various health life expectancy of WHO data set for years 2000 and 2015 on the world map by R using packages ggmap and ggplot2.
Cognitive Radio based on Wireless Broadband
Contributed towards the preliminary functions of Cognitive Radio, Cognitive Radio Security, and their usage on Wireless Broadband. In addition, proposed a MAC algorithm regarding the optimal allocation of users and spectrum efficiency. Published a couple of papers on Signal Processing and Cognitive Radio.
Analysis and Visualization of Airplane Safety data by Tableau
The project investigated the safety of airplanes for a newspaper article. To support this objective, data visualization tool named Tableau was used to explore the data and discover interesting facts thus creating relevant visualizations. Moreover, the report gives the analyses of the findings and visualizations by showing the significance of them along providing conclusions.
Exploratory Data Analysis with Python: Medical Appointments Data
Architected a zero-egress, multi-document enterprise RAG pipeline using Llama-3, ChromaDB, and LangChain. Implemented multi-query retrieval and independent RAGAS auditing to reduce hallucinations and achieve 50% guardrail adherence for fintech compliance. Listed as an open source contributor.
Volunteering
Volunteer
Publications
A review on security threats in Cognitive Radio
2014 4th International Conference on Wireless Communications, Vehicular Technology, Information Theory and Aerospace & Electronic Systems (VITAE) · May 12, 2014
2014 4th International Conference on Wireless Communications, Vehicular Technology, Information Theory and Aerospace & Electronic Systems (VITAE) · May 12, 2014
A Study on Mother-Tongue Interference in North Indian Spoken English