Summary

Overview

Work History

Education

Skills

Honors And Awards

Timeline

Anuvrat Kulkarni

AI Architect

Summary

Professional data scientist with strong background in statistical analysis, machine learning, and data visualization. Skilled in Python, Machine Learning, Generative AI and various data processing tools, with focus on delivering actionable insights. Known for collaborative approach, adaptability, and consistently achieving impactful results in dynamic environments. Recognized for problem-solving abilities and innovative thinking in leveraging data to drive business decisions.

Overview

years of professional experience

years of post-secondary education

Work History

Data Scientist Manager

EY GDS

01.2022 - Current

Generative AI - Application Development - Speech Bot Implementation: Refined a real-time chat solution using Azure OpenAI 4o Realtime and FastAPI WebSocket , enhancing user engagement by 50% and satisfaction scores by 60%, while facilitating over 2,000 monthly interactions and reducing instructor-led sessions by 30% across multiple training programs.
Generative AI - Application Development - Record Keeping Agreement : Spearheaded the design and implementation of a Generative AI solution using AWS Textract and Python , automating critical financial data extraction, achieving an 80% reduction in manual effort and 50% decrease in operational costs; enhanced data accuracy by 60% through FastAPI-driven validation processes, improved workflow automation by 60% with CI/CD and Docker , and ensured data integrity with proactive discrepancy identification.
Generative AI - Smart Call-Center Solution , Increasing Agent Productivity by 60%: Developed an LLM and data ingestion pipeline integrating Azure DevOps , LangChain , and ServiceNow to manage post-call actions, including summarization and ticket updates; utilized Azure speech-to-text for audio transcription while masking confidential information, handling 300-400 calls per center, boosting agent productivity by 60% and reducing average call handling time by 40%.
Transformers - Document Digitization for Netherlands-Based Client: Implemented a robust data extraction pipeline using Azure Form Recognizer to digitize and extract structured information from 5,000+ PDF formats, improving data accessibility by 40%; enhanced a BERT -based classification system for chemical information, boosting processing efficiency by 70%, and created a scalable solution with Python and Azure Functions, achieving a 50% reduction in processing time while handling over 10,000 documents monthly.
Generative AI - SDLC Asset Building : Assessed architecture diagrams for compliance, enhancing adherence to best practices by 30% across 15+ projects; conducted comprehensive code reviews to standardize pipelines, reducing bugs by 25%, and established detailed test cases for 100+ use cases, refining project delivery timelines by 20% through thorough testing and alignment with requirements.

Senior Data Scientist

Infosys Ltd.

06.2020 - 12.2021

PySpark + Machine Learning - Anomaly Detection in Retail: Utilized PySpark and Python for exploratory data analysis on 1 million+ retail records, improving model accuracy by 30% through feature extraction and outlier treatment; executed a Linear Regression algorithm for daily sales forecasting with 95% accuracy in anomaly detection, deployed on Azure Databricks to efficiently process large datasets and enhance data accessibility by 40%.
Python - Code Optimization and Infrastructure Cost Optimization : Optimized code and infrastructure using the multiprocessing module in Python to parallelize predictions on over 500 products, enhancing processing efficiency. Boosted execution speed by 90% while maintaining accuracy, resulting in a 70% reduction in operational costs on Azure Databricks.
Python EDA - Multi-Touch Attribution: Enhanced a Python framework for Multi-Touch Attribution (MTA) to analyze customer journeys for 10,000+ purchases, improving attribution accuracy by 25% through decay factor calculations across 5 digital media platforms; developed a dashboard to display state-wise sales anomalies ( ToBit Regression ), enabling real-time insights for 15+ regional managers.

Data Scientist

SG Analytics Pvt. Ltd.

12.2019 - 04.2020

NLP Developer - NLP Content Classification (End-to-End Product): Classified content and comments into specific categories using unsupervised learning techniques on 50,000+ entries, enhancing categorization accuracy by 30% through unique bigram identification; developed a user interface that minimized platform dependency and improved user experience for 1,000+ users using Python, NLP, HTML, CSS, JavaScript, jQuery, and Flask API.

Associate Process Manager

eClerx

05.2019 - 12.2019

Deep Learning Developer - Image Auditing Tool - Application Development: Enforced object detection and classification using Single Shot MultiBox Detector (SSD), enhancing accuracy by 40% and reducing inference time by 60% through GPU acceleration and TensorRT ; initiated MLFlow to optimize training parameters, cutting developer time by 70%; architected a modular RESTful API system with Flask , integrating techniques like blur detection, transfer learning , and OCR , facilitating collaboration among 5+ team members using Git .
MLOps - Automated Continuous Learning and Model Building: Developed a system to identify image attributes like blur, dimensions, and watermark presence, processing 10,000+ images; implemented Python scripts for automatic model retraining and validation, reducing human supervision by 70% through admin email notifications for results; monitored model performance using Weights & Biases , leveraging technologies such as TensorFlow , Flask, Git, and GPU acceleration .

Associate Software Engineer

Accenture Ltd.

12.2016 - 05.2019

Deep Learning Developer - Image Search Engine - Application Development: Generated an Image Search Engine using LSH and ANNOY for similarity searches on 100,000+ images, deployed as a microservice to enhance user experience for 1,500+ users; hosted on AWS (EC2) and utilized Git for version control, facilitating collaboration among 5+ team members; employed technologies including Python, Computer Vision , CNN , TensorFlow , Flask , and multithreading .
Natural Language Generation and Natural Language Understanding: Developed a natural language generation project from scratch, handling 5,000+ unique queries; utilized StanfordCoreNLP , NLTK , and Python with multithreading and GPU acceleration to improve processing speed by 50%; implemented POS tagging , NER , and sentence realization , creating a REST API in Java for linguistic typology, and enhanced deployment efficiency by 30% through end-to-end services and microservice deployment.

Education

Bachelor of Engineering - Information Technology

MIT College of Engineering

Pune

08.2012 - 06.2016

Skills

Experience: 84 Years of Experience in designing and developing Machine Learning, Deep Learning, Generative AI, and NLP, MLOps solutions for various industries including Retail, Banking, Telecommunications, and Chemicals, Wealth Asset Management

Management: Having experience in Solutioning, Architecting and Managing for greater than 3 years where I led 7 projects, primarily focusing on client communication while also performing an Individual Contributor role Also involved in the POC, POV, MVP and Initial Project Scope Understanding

Programming Language: Python, PySpark, Java, SQL, R, SAS

Frameworks: Transformers, LLM, VLM, PyTorch, LangChain, LlamaIndex, FastAPI, Scikit-Learn, SpaCy, TensorFlow, Keras, Huggingface, NLTK, TextBlob, BERT

MLOps: Weights and Biases, Databricks, MLFlow

LLM: OpenAI-4o, OpenAI-4o-Realtime, OpenAI, Anthropic Claude 35 Sonnet, Mistral

Platforms: Azure Cloud, AWS Cloud, Linux, Azure OpenAI, AWS Bedrock

Deployment: Kubernetes, Docker, GIT, Azure Function, Azure Web Services, AWS Lambda, Step Functions, ECR, ECS, DynamoDB, CosmosDB, CI/CD, Document Intelligence

Learning Agentic AI: AutoGen, LangGraph

Learning LLM Finetuning: LoRA, QLoRA, PEFT