Anurag Ghosh

MS in Artificial Intelligence

Janurary 2021 - Present

Relevant Coursework

• Machine Learning
• Large Language Models
• Pattern Recognition & Computer Vision

B.E. in Computer Engineering

July 2015 - May 2019

Relevant Coursework

• Database Management Systems
• Algorithms & Optimization for Big Data
• Machine Learning

Ribbon Communications

Januray 2025 - Now

Data Scientist

• Architected a generic end-to-end log analysis platform capable of ingesting raw, unstructured logs and transforming them into feature-enriched, clustered streams of correlated events. Conducted extensive literature review to integrate advanced techniques such as real-time log template extraction via online learning, unsupervised time-series correlation analysis, and BERTopic modeling for dynamic grouping of semantically related log events. The solution’s effectiveness in automated root cause analysis and event correlation directly contributed to securing a significant purchase order from Altice, expanding enterprise adoption of Ribbon’s analytics capabilities.
• Engineered a ML pipeline for optical performance monitoring, leveraging in-band power signal features to train quantile XGBoost and Gaussian Process regressors for Optical Signal-to-Noise Ratio (OSNR) estimation. Applied advanced statistical EDA and physics-informed feature engineering to mitigate data sparsity effects, resulting in a twofold increase in model predictive fidelity.
• Built an multimodal AI framework in Langflow that identifies statistical correlations among network failure events and employs an LLM with LlamaIndex-based retrieval-augmented reasoning to evaluate which correlations are logically and technically meaningful, providing interpretable causal insights and prescriptive remediation strategies.
• Built a chat-driven analytics platform for conversational EDA, automating chart/dashboard creation in Apache Superset using a multi-agent LlamaIndex workflow with a self-correcting SQL agent and MCP-exposed Superset APIs as agent tools.

Ribbon Communications

May 2024 - August 2024

Data Science and Machine Learning Intern

• Developed a URL categorization service utilizing a Selenium based web scraping engine. Leveraged a finetuned RoBERTa model to perform categorization on scraped webpage & data with a Zero Shot DeBERTa model for further sub categorization.
• Achieved a F1 score of 85% & reduced unknown/new URLs categorization & processing time on client (AT&T) side by 70%.
• Developed an automated log preprocessing and enrichment system leveraging template learning algorithms, NLP-based unsupervised clustering, and few shot severity classification to structure raw logs into event streams, highlight temporal correlations, and accelerate system troubleshooting and root cause analysis.
• Integrated preprocessing pipeline with a Streamlit app, allowing users to zoom into anomaly windows & analyze log clusters to identify potential trigger events. Enabled chat with logs of interest using an OLLAMA chatbot for further insights.

Universal Music Group

July 2023 - December 2023

Data Science Analyst Intern

• Designed LightGBM based Customer LTV Models for predicting purchase propensity & identifying superfans (high value) customers. Processed millions of online transaction data to train model, and interpreted predictions using Shapley. Elevated leadership’s awareness of data-driven approach to fanbase analysis & customer valuation leading to its prioritization for next quarter.
• Prepared a feature extraction module for raw emails of artist’s release campaigns. Devised a BERT based active learning technique to achieve agile labeling & training for models to tag emails with predefined categories & unsupervised labels. Achieved a 10% increase in performance of downstream in-production, engagement assessment models with the features.
• Developed a Bayesian Linear Regression model to predict artist’s revenue from factors like music, merch release & streaming data. Provided business with a Streamlit portal to tweak various factors & use the model to analyze their effect on the revenue.

WINES Lab Northeastern University

January 2023 - September 2023

Graduate Machine Learning Research Assistant

• Explored machine learning based radio fingerprinting by training neural networks like Alex net & Vision Transformers on sequences of in-phase (I) and quadrature (Q) data from LoRa devices to identify the devices in varying external scenarios.
• Achieved 99% accuracy across all device configurations when training and testing data was collected on the same day.

TIAA

July 2019 - July 2022

Analyst, Software Developer

• Lead a team of 4 developers overseeing development & production of 3 spring boot microservice applications.
• Independently supported & collaborated with business partners on production deployment for TIAA's retirement income evaluation appication.
• Achieved a 100% success rate with more than 20 releases within a year by making applications rapid release compatible with OpenShift integration. Reduced production deployment time from hours to a few minutes.
• Analyzed data and generated reports for insurance agents by creating KPI dashboards using Power BI. Streamlined user session and engagement tracking through automated SQL Batch Procedures, resulting in reduced efforts for the insurance team.

Languages and Databases

Frameworks

Tools