Data Analyst Scientist
Resume:
Rajesh Kumar
************@*****.***
Indian
Fort Lauderdale
GitHub
Visa: L2
Work Experience
Data Scientist
KPMG Global Services
03.2021 – 03.2024
Bangalore, India
Data Analyst
KPMG Global Services
09.2019 – 03.2021
Bangalore, India
Service Desk Analyst
EF Education First
01.2019 – 08.2019
Bangalore, India
Engagements
Multimodal RAG System, KPMG US LLP
•RAG system capable of multimodal i.e., text to image and/or image to text generation.
•Text and images were retrieved by Azure’s Document Intelligence along with its metadata i.e., page number, chunk ID.
•Captions for images were generated by utilising Azure’s Analyze Image API.
•Utilised OpenAI’s CLIP model for text-image pair training.
•Utilised ChatGPT 4 Vision for image and ChatGPT 4 for text generation.
•Based on PyTorch’s cosine similarity we extracted nearest neighbors for user query. Embedding were stored in Pandas DataFrames during POC.
•Language: Python
•Dev Environment: Jupyter Notebook, VSCode
Research Digest, KPMG US LLP
•Developed prototype using Open source Llama models for building the RAG system and hosted on private cloud environment.
•Tool for researchers and healthcare professionals to query and retrieve summarised journals from PubMed/National Library of Medicine database.
•Query PubMed database based on persona, timeline and keywords. Input title-abstract to GPT3.5 Turbo to classify high priority articles with justification and then summarize them using ChatGPT 4 Turbo.
•Enabled intelligent QnA on articles using GenAI's embedding and FAISS vector database.
•Language: Python
•Front-end: Streamlit
1 / 2
Object Detection, KPMG US LLP
•Detect trash in ATMs such as paper, water bottles
•Crowd counting (footfall) in shopping complex or malls
•Algorithm: YOLOv3 and YOLOv4
•Language: Python
Digitalised Vendor Risk Assessment Chatbot, Comcast LLP
•Developed virtual agents on ServiceNow to automatically onboard vendors, collect vendor information and analyze risks associated with vendor
•The digital worker replaced the manual process by inculcating a rule based mechanism which eradicated human errors and biases
•The digital worker is used by hundreds of employees and vendors across the globe
•Language: ServiceNow Scripting
Kacey Conversational AI, KPMG US LLP
•Developed intelligent Conversational AI Assistants using Natural Language Processing (NLP)
•Utilised pretrained Predictive Intelligence model and Similarity Indexing
•Framework: Azure Bot Framework (LUIS), RASA
•Language: Python
Education
Master of Science in Artificial Intelligence and ML, Liverpool John Moores University
2020 – 2021
Post Graduate Diploma in Artificial Intelligence and ML, International Institute of Information Technology - Bangalore
2019 – 2020
Bachelor of Engineering in Information Science and Engineering, SJC Institute of Technology
2013 – 2018
Certificates
Python for Data Science
LinkedIn Learning
SQL Essential Training
LinkedIn Learning
Google Dialogflow
Google Cloud
Alteryx Machine Learning
Alteryx
2 / 2
Contact this candidate