article Publications

Research &
Publications.

Peer-reviewed work at the boundary of machine learning, environmental data science, and explainable AI.

I.

Published papers.

Research / 01
InJET · Vol. 16, No. 2 December 24, 2025

Comparative Analysis of Traditional and Ensemble Models for Water Quality Index Prediction with Explainable AI

Amrit Kandel, Rajad Shakya

Applied comprehensive EDA and preprocessing over EPA Ireland coastal water monitoring data to predict the CCME Water Quality Index (CCME-WQI). Built and compared multiple models including Linear Regression, Random Forest, and XGBoost (achieving R² = 0.991). Used SHAP (SHapley Additive exPlanations) analysis for model interpretability and feature importance, contributing to transparent, data-driven environmental decision making.

Machine Learning XGBoost SHAP Environmental AI R² = 0.991
DOI: 10.3126/injet.v16i2.59782  ·  Read paper ↗
II.

Ongoing work.

In Progress / 02
WIP-01 In Progress

End-to-End Person Video Retrieval Using Track-Level Robust Scoring and Dynamic Thresholding

IEEE conference manuscript under development. Explores multi-camera person re-identification combining YOLO detection, ByteTrack, OSNet embeddings, and a novel track-level scoring system for robust retrieval.

Person Re-IDTrackingComputer Vision
WIP-02 Exploring

AI-Enhanced Research Paper Analysis

Building an intelligent research assistant that analyzes PDFs, generates contextual summaries, extracts key contributions and research gaps, and enables semantic search across related literature using vector embeddings.

LangChainRAGNLP
WIP-03 Learning

Azure Cloud ML & Data Engineering

Exploring the full Azure AI ecosystem — integrating Azure ML, Databricks AutoML, and Cognitive Services with end-to-end data pipelines from ADF ingestion to Power BI visualization.

Azure MLDatabricksCloud
III.

Research interests.

Focus areas / 03
visibility

Computer Vision

Object detection, tracking, person re-identification, multi-camera systems, and video understanding pipelines.

psychology

Explainable AI

Making black-box models interpretable — SHAP, LIME, feature attribution, and model transparency for real-world trust.

hub

Applied ML

Bridging research and engineering — deploying ML systems at scale, from Jupyter notebooks to production APIs.

eco

Environmental Data Science

Applying ML to environmental monitoring — water quality, climate data, and sustainability-focused AI solutions.