My Projects
A showcase of AI/ML solutions across healthcare, finance, and logistics
Featured Project
EEG Cognitive State Classifier
Deep learning model for real-time stress and fatigue detection from EEG data, deployed in ICU monitoring devices with >90% precision.
View ProjectAll ProjectsHealthcare AINLP & FinanceLogisticsFeatured
EEG Cognitive State Classifier
Deep learning model for real-time stress/fatigue detection from EEG data in ICU environments.
PyTorchONNXSignal ProcessingPlotly+4
- Achieved >90% precision in clinical simulations for stress/fatigue detection
- Optimized inference speed to <200ms latency on edge devices through model pruning
Brain MRI Segmentation Pipeline
U-Net-based deep learning tool for automated segmentation of brain anomalies in MRI scans.
TensorFlowOpenCVDICOMAWS+4
- Achieved high Dice scores on tumor segmentation tasks
- Implemented DICOM image loaders with advanced preprocessing pipeline
Edge AI Deployment for ICU Monitoring
Optimized neural network model deployment to resource-constrained embedded devices in ICU environments.
PyTorchONNXEmbedded LinuxDocker+3
- Achieved sub-200ms inference time on resource-constrained edge devices
- Reduced model size by 70% through quantization-aware training and pruning
Email Intent Classification Tool
NLP-based email triage system using fine-tuned BERT to automatically classify support inquiries.
PythonHuggingFaceBERTFastAPI+3
- Saved 30+ hours/week in manual email triage time
- Implemented feedback loops to improve model with user corrections
Loan Default Explainability Layer
SHAP and LIME powered insight engine for explaining machine learning loan risk predictions.
SHAPLIMEScikit-learnXGBoost+3
- Enhanced model transparency for financial analysts and compliance teams
- Designed intuitive UI widget for visualizing feature contributions to risk scores
ETA Prediction Engine
ML-based delivery time estimation system that reduced ETA error from 30min to under 10min.
PythonScikit-learnGoogle Maps APIGPS Data+3
- Reduced ETA error margins from 30 minutes to under 10 minutes
- Developed hybrid ML + ruleset approach that adapts to traffic anomalies
Geospatial Route Clustering Engine
Optimized delivery routing system using density-based clustering for more efficient dispatch planning.
PythonOpenCVHDBSCANNode.js+3
- Minimized total travel distance while respecting delivery constraints
- Reduced dispatching time and fuel consumption for major logistics clients
Truck Demand Forecasting System
LSTM-based time-series predictor for regional truck capacity needs, improving fleet utilization by 18%.
PythonKerasLSTMAWS Lambda+3
- Improved fleet utilization by 18% through accurate capacity forecasting
- Handled multiple time-series sources with varied granularity and patterns
Hospital Readmission Risk Model
Predictive model to estimate 30-day readmission risk for discharged patients using EHR data.
PythonScikit-learnPandasNumPy+3
- Created standardized feature extraction for inconsistent healthcare data
- Developed interpretable models with strong predictive performance
Interactive AI Demos
Experience interactive versions of these projects in the AI Lab. Try out working demos with simulated data.
Visit the AI Lab