01 / Frame
Problem first thinking
Define the question, users, evidence, and success measure before choosing the tools.
From complex data to clear decisions.
I build auditable AI workflows, predictive models, and interactive analytics that turn difficult questions into tools people can use. Explore the methods, evidence, and decisions behind the work.
Selected work
Real problems. Traceable results.
Selected projects across computer vision, healthcare operations, and AI assisted research. Each shows the question, the method, and what the evidence supports.
Explore the full project library
Browse the wider collection by problem area, method, or tool.
01 / The approachFrom problem to proof
About me
The question comes first. The technology follows.
I begin with the decision someone needs to make, then work backward to the evidence. That might call for a model, data pipeline, dashboard, or application, but the method should serve the problem, not the other way around.
I take projects from exploration to a documented, working result. Across machine learning, computer vision, local AI, SQL analytics, and forecasting, the through line is validation: traceable measurements, reproducible methods, and clear communication about what the evidence supports.
02 / Validate
Reliable AI and data
Test the output, show the assumptions, and make the limits of a result visible.
03 / Deliver
Build, document, deliver
Turn the analysis into a working tool or clear report that others can inspect and use.
02 / CapabilitiesMethods chosen for the work
Skills
A toolkit for the whole journey.
From building a trustworthy data foundation to testing a model and making the result usable, I bring the right methods together around the question.
AI and machine learning
Classification, regression, computer vision, NLP, recommendation systems, anomaly detection, and rigorous model evaluation.
scikit-learnPyTorchU-Net
Local AI and intelligent workflows
Local LLMs and vision language workflows, evidence grounded extraction, structured outputs, routing, planning, and human review.
Local LLMsStructured JSONWorkflow Logic
Data systems and analytics
Python and SQL pipelines, data cleaning, validation, metric design, warehouse style modelling, and operational dashboards.
PythonMySQLData Validation
Decision intelligence
Customer, revenue, healthcare, policy, operations, and behavioural analysis translated into clear, decision ready findings.
EDADashboardsReporting
Forecasting, risk and experimentation
Time series forecasting, risk scoring, uncertainty, rolling backtests, leakage controls, anomaly detection, and model comparison.
SARIMAXRisk ModelsEvaluation
Delivery and reproducibility
Interactive Streamlit apps, reproducible notebooks, documented repositories, analytical reports, and traceable end to end workflows.
StreamlitGitHubDocumentation
