MZQMohd Zamin Quadri

Research

Research

My work asks how a machine learning model can report more than a single answer: where it is likely to be wrong, whether its stated confidence can be trusted, and when a prediction should be sent to a person instead.

RecruiterWhat did the work establish?
EngineerHow does uncertainty change a system decision?
Researcher / studentWhich protocol supports each claim?
Research focusPrimary recordSupporting work

01Research focus

What I work on

Reliable machine learning and uncertainty on graph models are the main body of work. Scientific modelling supports it. Identifiability is an interest rather than a result I claim.

Primary research

Reliable machine learning

Uncertainty quantification for graph surrogates: ranking likely error, calibrating uncertainty, constructing conformal intervals, and routing uncertain predictions to review.
  • Uncertainty Quantification
  • Calibration
  • Conformal Prediction
  • Selective Prediction
  • Graph Neural Networks
Uncertainty Quantification for Traffic Prediction

Emerging inquiry

Mathematical structure of learned models

Neural-network identifiability is an active mathematical ML direction, presented as an area of inquiry rather than a completed research result.
  • Neural Network Identifiability

02Primary record

Uncertainty Quantification for ML Models in Transportation Policy Analysis

My completed M.Sc. thesis. A graph neural network stands in for a traffic simulation of Paris, and the work is about making its confidence trustworthy: how accurate it is, whether its uncertainty ranks its own errors, whether its stated ranges hold, and when to set a prediction aside.

Academic research / Master's thesis submitted

Uncertainty Quantification for Machine Learning Models in Transportation Policy Analysis

Built on the MATSim corpus and PointNetTransfGAT infrastructure of prior work, then evaluated training variants, MC Dropout, deep ensembles, sigma scaling, split and adaptive conformal prediction, selective prediction, CQR, and error-detection diagnostics.

Evaluated scope
100 scenarios3,163,500 link predictions
Uncertainty ranking
rho 0.482MC uncertainty vs. absolute error
Selective review
41.2%Lower MAE at 50% retention
Explore the thesis researchEngineering case studyCanonical repository

03Supporting work

Supporting work

Smaller projects around the same subject. They are not presented as equivalent to the thesis.

Group coursework

Uncertainty Quantification in Hydrology

A team seminar connecting HBV rainfall-runoff calibration, local and global sensitivity analysis, and input/output uncertainty propagation.Group coursework; individual ownership of each result is not established

Synthetic demonstration

Synthetic Streamflow Forecasting Benchmark

A controlled benchmark that scores ridge, XGBoost and SARIMAX against the baselines that actually matter on a daily series, and finds that the gradient booster loses to predicting yesterday's value.Synthetic data cannot establish real-catchment performance

04Learn

Turn the research question into an operational concept

The learning layer explains one decision mechanism at a time, while the thesis record preserves protocols, evidence status, and limitations.

tutorial4 min read

Calibration Is Not Classification Accuracy

A model can be accurate and badly calibrated, or well calibrated and useless. What each property means, how to measure both, and why thresholds are a decision.

Machine Learning
tutorial5 min read

Production AI Systems Need Refusal Paths

Every stage that can degrade should be able to decline. Where refusal belongs across ingestion, retrieval, extraction and inference, and what omitting it costs.

Reliable AI Systems