Aswini Sivakumar
PhD Student in Computer Science · Oakland University, Rochester MI
About
I am a PhD student in Computer Science at Oakland University, working at the intersection of computer vision, autonomous systems, and vision-language models. My research focuses on building systems that can perceive, reason, and act in complex real-world environments using retrieval-augmented generation and contrastive learning.
News
| Date | Update |
|---|---|
| 12.2025 | Won SAS Hacakthon 2025 in two tracks - SAS Viya Workbench and Sustainability |
| 01.2025 | Started my PhD at Oakland University |
| 11.2024 | Won SAS Hacakthon 2024 in Energy track |
| 03.2024 | Presented at the Interdisciplinary Applications of AI and Data Analytics Symposium - “Use of Chatbots in Medical Education,” Oakland University |
| 06.2023 | Won SAS Hacakthon 2023 in Forecasting track |
Research Interests
- Autonomous Driving - End-to-end autonomous pipelines with multi-modal perception
- Vision-Language Models - Fine-tuning and adapting large VLMs for scene understanding, action prediction, and justification generation
- Retrieval-Augmented Generation - Hybrid retrieval and evaluation frameworks
- Computer Vision & NLP - Object detection, semantic embeddings, sentiment analysis, and biomedical text mining
- AI Safety & Robustness - Safety in autonomous systems
- Applied ML - Predictive modeling for healthcare, energy forecasting, and business analytics
Education
Ph.D. in Computer Science Oakland University · Rochester, MI · 2025 – Present
Working under the supervision of Dr. Yao Qiang in the Secure, Aligned, Fair, and Ethical (SAFE) AI Research Lab.
M.S. in Business Analytics Oakland University · Rochester, MI · 2022 – 2024
Coursework in machine learning, AI ethics, data-driven decision making, and statistical modeling.
M.B.A. in Finance BITS Pilani · India · 2019 – 2020
Specialization in corporate finance, capital budgeting, and quantitative business analysis.
B.Tech in Computer Science and Engineering Amrita University · India · 2011 – 2015
Foundation in algorithms, software engineering, and computer systems.
Experience
Graduate Research Assistant Oakland University · Rochester, MI · 2023 – Present
Associate Engineer Cognizant · India · 2018 – 2020
System Engineer Tata Consultancy Services · India · 2015 – 2018
Projects
Medical RAG Evaluation Framework
RAGXMedicalQAEvaluator with novel Context Utilization Efficiency (CUE) metrics. Hybrid BM25 + dense retrieval using MedCPT encoders, evaluated on PubMedQA and USPSTF clinical guidelines with DeepSeek-V3 as LLM judge.
MedCPT BM25 Weaviate DeepSeek-V3 PubMedQA
Cancer Subtype Classification
ML pipeline for endometrial cancer subtype prediction on the TCGA/UCEC dataset. Handles class imbalance via SMOTE and compares six ML models with an emphasis on parsimonious Decision Trees.
SMOTE Random Forest TCGA Scikit-learn
Explainability on transformer-based models
Fine-tuned DistilBERT on a restaurant reviews dataset for sentiment analysis, then benchmarked three post-hoc explainability frameworks — SHAP, Transformers-Interpret, and Ferret XAI. An interactive Gradio dashboard visualizes feature attributions in real-time.
DistilBERT SHAP Ferret XAI Gradio Transformers Explainability
LinkedIn Profile NLP Classifier
Classifying professional profiles using natural language processing. Feature extraction from unstructured text with classical ML models.
NLP XGBoost Python Scikit-learn
Energy Demand Forecasting
Predictive modeling for electricity demand using time series analysis and machine learning, applied to utility grid planning and demand response.
Time Series Scikit-learn Pandas Forecasting
Housing Buy vs. Rent Decision Tool
Data-driven decision support tool for housing market analysis, incorporating mortgage rates, opportunity cost, and long-term financial projections.
Finance Python Data Analysis