Aspiring Analytics Professional with ML Research Experience

Swarnadwip Bhattacharya

Pittsburgh, PA 412-***-**** ********@******.***.*** LinkedIn

EDUCATION & CERTIFICATIONS

Carnegie Mellon University Pittsburgh, Pennsylvania Master's in Business Analytics Anticipated: May 2026

● Relevant Courses: Machine Learning for Business Applications, Data Exploration & Visualization, Statistical Foundations of Business Analytics, Programming in R & Python Kolkata, India

Graduation Date: Jun 2024

● Publications: “Unlocking Domain Specificity: Fine-Tuning Llama-2 for Custom Datasets,” IEMTRONICS London, UK, 2024 (Best Paper) Published with Springer Journal. Indexed in Scopus. Certifications

● AWS Cloud Practitioner Essentials, Amazon Web Services, April 2023

● High Dimensional Data Analysis, edX, April 2021

WORK EXPERIENCE

VECC (Bhabha Atomic Research Centre)Kolkata, India AI/ML Research Intern Dec 2023 - Feb 2024

● Built a RAG system (FAISS + LangChain + Llama-2), achieving F1 = 0.83 and ROUGE-1 (+3.45%), ROUGE-2 (+6.38%), ROUGE-L (+0.17%), ROUGE-Lsum (+4.81%), improving output relevance and coherence.

● Released a cleaned, labeled corpus on Hugging Face with datasheets and eval scripts to support reproducible benchmarking and research collaboration.

● Implemented an evaluation harness (ROUGE, BERTScore) with time-split cross-validation, preventing leakage and enabling ablation studies.

National University of Singapore Singapore

ML Research Intern Jul 2022 - Sep 2022

● Built a real-time facial emotion classifier (Python, TensorFlow/Keras, OpenCV), improving accuracy from 60% 85% across diverse expressions.

● Designed an augmentation pipeline (flips, crops, blur, CLAHE), expanding the dataset from 2K 10K images and boosting robustness on hard-to-distinguish emotions (surprise vs fear).

● Optimized preprocessing workflows (denoising, CLAHE, Gaussian blur), reducing noise, improving precision, and ensuring real-time efficiency on CPU. PROJECT EXPERIENCE

Financial News Sentiment Analysis

● Fine-tuned a finBERT model on financial headlines (Reuters/FinViz); on an out-of-time test set (N=10,000) achieved macro-F1 = 0.89 and AUC = 0.94, accurately predicting stock sentiment.

● Engineered a daily inference pipeline (Python, scikit-learn) and Tableau dashboard, generating trading signals with a backtested Sharpe ratio of 1.21.

Fraud Detection System

● Built a fraud detection model on 6M+ transactions (0.13% fraud rate) using scikit-learn; achieved 94% accuracy with 91% recall, capturing most fraudulent activity.

● Deployed as a Streamlit web app, using time-based splits and leakage checks; SHAP analysis highlighted key fraud drivers (transfer, cash_out) to inform rule updates. SKILLS & INTERESTS

Skills: Programming & Data Tools: Python, R, SQL, Tableau, OpenCV, AWS, GCP ML & DL Frameworks: TensorFlow, Keras, scikit-learn, Hugging Face Techniques: Machine Learning, Deep Learning, NLP, Predictive Modeling, Sentiment Analysis, Data Augmentation, Logistic Regression, Model Fine-tuning, Data Visualization Interests: Animal Welfare, AI Research, Financial Markets, Data Science, Community Service

Contact this candidate