EstimateRUL Predictive Maintenance
Built an end-to-end Remaining Useful Life (RUL) estimation application on NASA CMAPSS FD004 datasets using XGBoost. Supported by a FastAPI backend, RBAC security patterns, and real-time sensor inferences.
Aspiring data scientist skilled in machine learning pipelines, predictive modeling, multimodal AI research, and deploying production-ready ML applications.
I am an Information Technology student currently pursuing my undergraduate degree from NIST University and aspiring data scientist with a deep interest in machine learning pipelines, predictive modeling, and multimodal AI research. I enjoy building systems that process complex datasets and deliver actionable insights.
My background includes hands-on experience in building multimodal benchmarks for Vision Language Models, developing predictive maintenance systems, and designing context-aware student query assistants using low-resource Small Language Models.
Gained hands-on training in Salesforce fundamentals, including data modelling, security models, generative AI, and modern CRM systems. Learning to build custom objects, page layouts, workflows, Apex programming, and Lightning components.
Built EduVLM-Bench, a multimodal VLM evaluation benchmark for prerequisite detection. Simulating student error responses with Gemini 2.0 + Gemma3 4B. Evaluated 9 VLMs, with Qwen 2.5 VL achieving 38.5% accuracy. Deployed a live web platform: eduvlm-hj3z.onrender.com.
CGPA: 8.97 / 10.0
Grade: A
Grade: A+
Led multiple university workshops and inter-campus technical programs based on new frameworks with 150+ attendees. Directed teams of coordinators to design and deploy hands-on bootcamps with 40+ student memebers of our club.
Comparative study of Tiny Language Models and fine-tuning approaches for context-aware student query resolution. Explores efficient indexing pipelines to improve factual recall and query latency in low-resource environments.
Awarded Prestigious Research Fellowship by the Department of Science and Technology, Government of India. Fully Sponsored Research, Focused On Evaluation and Benchmarking Of LLMs & Multimodal Vision Language Model (VLM).
Built an end-to-end Remaining Useful Life (RUL) estimation application on NASA CMAPSS FD004 datasets using XGBoost. Supported by a FastAPI backend, RBAC security patterns, and real-time sensor inferences.
Designed low-resource conversational AI leveraging Tiny Language Models (77M–1.7B params) and fine-tuned RAG pipelines. Achieved 81.8% semantic accuracy retrieving curriculum notes.
Developed a Decision Tree ensemble classifier predicting churn risk profiles based on usage metrics. Includes feature importance explanations and an interactive Streamlit inference dashboard.