Severity-Weighted Calibration Error for Reliability Assessment Under Outcome-Severity Imbalance
Aggregate calibration scores mask failures in high-acuity subgroups. Per-band isotonic recalibration consistently fixed this; global recalibration did not.
Aman Chandra H, K.C. Narendra · 2026
Intravenous Bag Monitoring and Alert System — A Prototype
Real-time IV fluid monitoring using microcontroller-based non-contact capacitive sensors, integrated with an IoT platform for centralised alerting. Presented at ICBEST 2025.
Aman Chandra H et al. · 2025
Severity-Aware Evaluation of ICU Mortality Prediction Using Risk-Weighted Error Metrics (RWEM)
Introduces RWEM — a metric that penalises missed deaths by patient severity. Evaluated across 5 models and 3 critical care databases.
Aman Chandra H, K.C. Narendra · 2025
Prevalence Shift and ICU Mortality Miscalibration Across Institutions
Zero-label Bayesian correction for prevalence shift achieving 56–65% calibration error reduction across 185 hospitals without model retraining.
Aman Chandra H, Stuti Shivhare, Suyash Sahu · 2026
Language as a Hidden Variable: Measuring Behavioral Divergence in Multilingual Large Language Models
A controlled multilingual study introducing BDS and showing cross-language behavioral divergence up to 2.5x within-language variation.
Aman Chandra H, Sameer Kulkarni, S Nithin, Khushboo Pandey · 2026
NOMOS: RAG Framework for Jurisdiction-Specific Indian Legal QA
Privacy-preserving legal QA over 2.32M characters of Indian statutory text using local LLM inference, ChromaDB, and React + Firebase. Outperformed GPT-4 on citation accuracy.
Sameer Kulkarni, Aman Chandra H et al. · 2025
Clinical Risk-Weighted Evaluation of ICU Mortality Prediction Models (CRWS)
A cost-sensitive evaluation framework for ICU mortality, penalising missed deaths over false alarms. Evaluated across 5 model classes on 3 independent critical care databases.
Aman Chandra H, Abhijna Shivaprakash et al. · 2025
Does Model Compression Amplify Clinical Bias? A Subgroup Fairness Audit of Quantized ICU Risk Models on MIMIC-IV
Audits whether INT8 quantization, pruning, and knowledge distillation amplify subgroup fairness gaps in ICU AKI prediction — and uncovers a quantization failure mode caused by outlier EHR values.
Aman Chandra H · 2026
Graph-Based Polypharmacy Risk Networks for Adverse Event Prediction in ICU Patients: A MIMIC-IV Analysis
Compares graph embeddings against tabular features for predicting AKI, delirium, and bleeding from ICU polypharmacy patterns — and catches a label-leakage bug that was inflating results before correction.
Aman Chandra H · 2026