Data Scientist II

HPE provides edge-to-cloud solutions, including computing, storage, networking, AI-driven services, and a range of enterprise servers across multiple generations.

At HPE, I designed and deployed real-time classification models including LightGBM, Random Forest, and XGBoost to categorize incoming customer cases, which reduced manual triage time by approximately 45% and improved prioritization accuracy to around 92%. This involved extensive feature engineering, including dimensionality reduction through PCA on both textual and categorical metadata. I also utilized Impala SQL and Dataiku DSS to preprocess over 10,000 customer case logs weekly and applied NLP models such as Word2Vec and GloVe for contextual text embedding and pattern recognition in customer feedback. To uncover latent trends in unlabeled server issue data, I conducted unsupervised learning using clustering algorithms such as KMeans, DBSCAN, and Hierarchical Clustering. Analyzing over 2.5 million server log entries, I identified six key failure clusters that enabled the development of automated alerts and contributed to a 28% reduction in unresolved server anomalies. I also built LLM-based tools using Anaconda on NVIDIA L40 GPUs to assist support teams in product inquiries and case resolutions. Using LangChain with Azure OpenAI API, I implemented retrieval-augmented generation (RAG) pipelines for question answering, while fine-tuning models like Flan-T5 for sentiment analysis and Gemma, Granite, and Llama for summarization and context retrieval, improving internal resolution rates by 73% across four departments. Additionally, I designed rule-based and statistical models to detect early indicators of customer issues, applying the Pareto principle to optimize prioritization and reducing detection time by 34%. I managed data pipelines integrating inputs from over six sources using Dataiku, ensuring data consistency and version control. To support data-driven decision-making, I developed and maintained three interactive Power BI dashboards tracking case inflow, model health, and team productivity. As the point of contact for four cross-functional teams, I led several generative AI initiatives covering summarization, sentiment analysis, question answering, and context extraction to enhance support operations and reduce external tool dependency.

Data Analyst

HSBC (Hongkong and Shanghai Banking Corporation) is a global financial services company offering banking, wealth management, and investment solutions across multiple markets.

At HSBC, I worked as a Data Analyst focused on trade and transaction data, where I extracted and processed over 20 million records daily through secure API pipelines into Data360 ETL workflows. I performed extensive EDA using Python and SQL, applying over 70 internal validation rules to enhance data quality and ensure regulatory compliance—improving accuracy metrics by 35%. I built a regression-based anomaly detection model that reduced manual exception handling by 2–3 hours/day and increased risk flagging efficiency by ~50%. I also developed and deployed classification models (Random Forest, XGBoost, Logistic Regression) using PCA-transformed trade metadata, achieving 90%+ accuracy in transaction categorization. Leading a team of 3 analysts, we automated critical monitoring workflows, improving system reliability and reducing script failures by 40%. Additionally, I created 5 MySQL summary tables for downstream QlikSense dashboards used by 3 key business teams and optimized modular Python ETL scripts, cutting processing time by ~40%. My role involved close collaboration with stakeholders to deliver actionable insights and accelerate data-driven decision-making.

Data Analyst Intern

MIQ Digital is a programmatic media company that provides data-driven advertising and analytics solutions for targeted marketing.

As a Data Analyst, I worked on programmatic advertising data for four key clients, leveraging Python-based data analysis techniques to generate weekly and monthly performance reports. This work led to a 15–20% improvement in campaign efficiency. I extracted and stored large-scale advertising data using DSP platforms and Amazon S3, and conducted in-depth exploratory analysis on Databricks using PySpark and SQL, handling over 10 million records per week. To streamline operations, I automated data pipelines with reusable scripts, reducing reporting time by 30–40%. I created actionable visual insights using Power BI and Python libraries like Seaborn and Matplotlib, highlighting key trends in CTR, CPA, spend, and audience behavior. Additionally, I curated clean datasets for internal teams on S3 to support faster decision-making and performed variance and anomaly analysis that directly contributed to data-driven business optimizations and improved ROI.

Masters in computer applications

Course Duration: Three Years.

Grade: First class distinction.

Bachelors in Physics Honors

Course Duration: Three Years.

Grade: First class.