Navika Maglani Data Scientist | ML Engineer | AI Engineer | Python, SQL, AWS |

A

AI-Powered Supply Chain Disruption Intelligence System

May 2026

Developed an AI-driven operational intelligence platform designed to predict supply chain disruption risk using machine learning, live weather intelligence, and global trade analytics. Integrated U.S. Census International Trade APIs, NOAA weather alerts, country-to-port dependency mapping, and time-series trade analytics into an interactive Streamlit dashboard for real-time disruption monitoring and simulation. Built and evaluated Logistic Regression and Random Forest models on 18,000+ international trade records across 250+ countries and regions, achieving ~81% accuracy and ~0.88 AUC for disruption risk classification. Incorporated SHAP explainability, class balancing techniques, rolling trend analysis, and operational dependency modeling

A

AI-Powered Portfolio Risk Intelligence Platform

Apr 2026

Developed an AI-powered portfolio risk analytics platform focused on explainable machine learning, financial forecasting, and real-time market intelligence. Built ML-based portfolio risk classification models and integrated SHAP explainability to identify feature-level drivers behind changing risk predictions. Implemented Monte Carlo simulations for forward-looking portfolio forecasting and incorporated technical indicators including RSI, MACD, Bollinger Bands, and rolling volatility analysis. Performed benchmark analysis against SPY using Sharpe Ratio and Maximum Drawdown metrics. Integrated NLP-based financial news sentiment analysis to evaluate market sentiment alongside technical and quantitative indicators. Designed an interactive Stre

M

MCP-Based Agentic AI System for Cleantech Retrieval & Reasoning

Oct 2025 - Nov 2025 (2 months)

Built an end-to-end Agentic AI system using the Model Context Protocol (MCP) to perform retrieval, reranking, summarization, and reasoning over 20,000+ cleantech media articles. Designed a modular multi-tool architecture integrating ChromaDB vector search, cross-encoder reranking, LLM-based grounded summarization, and explainable reasoning workflows. Key contributions: • Indexed 20,111 cleantech articles in ChromaDB using MiniLM-L6-v2 embeddings with metadata-aware retrieval • Implemented reranking using ms-marco-MiniLM-L-6-v2 to improve relevance beyond pure vector similarity • Built MCP tools for retrieval, summarization, reasoning, and full pipeline orchestration • Developed grounded answer generation with support classification (support

A

Automated Brain Tumor Classification & Segmentation using Hybrid Image Processing & CNNs

Sep 2025

Developed an end-to-end medical imaging pipeline for brain tumor segmentation and classification using T1-weighted contrast-enhanced MRI scans. Combined classical image processing techniques with deep learning to identify and classify meningioma, glioma, and pituitary tumors from MRI slices. Key contributions: • Built a hybrid segmentation pipeline using median filtering, histogram normalization, percentile thresholding, and morphological operations • Developed a custom 3-layer CNN architecture for tumor classification with 98% test accuracy • Implemented Grad-CAM++ visualizations to improve model interpretability and verify tumor-focused attention regions • Performed texture analysis using GLCM features for enhanced image characterization

F

Flight Delay Prediction (AWS-based Project)

May 2025 - Jun 2025 (2 months)

Constructed an AWS EMR pipeline with PySpark to process 1M+ flight records, predicting delays with 77%accuracy. Stored data in S3, queried with Athena, and built interactive dashboards for stakeholders.

C

Chicago Crime Data Analysis (Tableau Dashboard)

Feb 2025 - Mar 2025 (2 months)

Visualized crime trends from 2019–2024 using Tableau, heatmaps, clustering, and time-series analysis. Created interactive dashboards enabling users to explore crime trends by time and location.

D

DePaul chatbot (LLM-powered university Q&A system

Feb 2025

Designed an LLM-based chatbot using Llama for real-time data retrieval from DePaul’s webpages handled data preprocessing and integration for the Academic Calendar module.

F

Fashion MNIST Image Classification using Transfer Learning with ResNet50

Sep 2024 - Nov 2024 (3 months)

Fine-tuned a ResNet50 model pre-trained on ImageNet for Fashion MNIST, achieving 91.75% validation accuracy and showcasing the effectiveness of transfer learning. Deployed a production-ready Dash web application enabling real-time image classification and prediction explanation.

P

Personalized Content Recommendation Engine (Social Platform Simulated)

Apr 2024 - Jun 2024 (3 months)

Architected a recommendation engine using item-based collaborate filtering with cosine similarity and Pearson correlation. Increased recommendation accuracy by 20% through fine-tuning similarity metrics. Conducted trend analysis on user behavior to deliver personalized recommendations, improving user engagement by 15%.

B

Basketball Analysis (Data Analysis and Regression)

Jan 2024 - Mar 2024 (3 months)

Analyzed 500+ player performance datasets, achieving an R² of 0.75 in scoring efficiency predictions. Created 10+ data visualizations using Matplotlib and Seaborn to present insights to peers and instructors.

3

30-day Emergency Department Patient Readmission Prediction

Analyzed data from a Chicago based Healthcare Institute to identify Social Determinants of Health affecting patient readmissions. Leveraged various NLP techniques to extract meaningful features from the patient conversations and integrate with Demographics to improve patient readmission predictability.

Navika Maglani

About

Detail

Timeline

Résumé