Master's in Data Science @ UT Arlington

I work across data analytics, analytics engineering, data governance, and AI agent governance. Most of my projects start from the same question:

Can this data, metric, workflow, or AI agent actually be trusted enough to support a real decision?

Not "does the dashboard look right." More like, does anyone agree on what the metric means, who owns the definition, and what happens when the number changes? That is usually where things get interesting.

I am finishing my Master's at UT Arlington and looking for internship, co-op, or new-grad roles in data analytics, BI, product analytics, analytics engineering, or data/AI governance.

My strongest fit is with teams where reliable data and governed workflows are treated as real engineering problems, not afterthoughts bolted on after the dashboard ships.

These are the governance-adjacent projects I've contributed to. All three are verifiable.

Contributed to a curated collection of frameworks, tools, standards, and research focused on governing AI agents. The work connects directly to what I spend most of my time thinking about: how AI agents, data access, observability, and governance controls come together in production systems.

Invited as an early contributor to the AgentTrust repository. The problems here are the ones I keep coming back to: how should AI agents prove trustworthiness? What evidence should exist around agent workflows? How does governance move from documentation into something operational?

Opened a governance and security proposal in the A2A ecosystem — focused on vendor-neutral documentation for governance and security expectations in agent-to-agent communication. The core interest: how agents should follow rules, access data, and leave behind reviewable evidence.

A governance-ready analytics engineering project built around a simple idea: analytics systems should not only produce metrics, they should also prove whether those metrics are trustworthy.

The project includes reliable data pipelines, modeled analytics layers, data quality checks, governance marts, and decision-ready outputs using dbt, PostgreSQL, and SQL. The kind of system where if someone asks "can we trust this number?" there is an actual answer, not just a shrug.

A conceptual governance platform direction I'm developing. The question it tries to answer:

Can you show what your AI system did, what data it used, and whether it followed the right rules?

The design combines data contracts, agent audit trails, synthetic data generation for privacy-safe sharing, and compliance artifacts that turn system behavior into reviewable evidence. This is early-stage thinking, not a shipped product, but it represents the direction I'm building toward.

Projects focused on connecting metrics, experiments, and product decisions. The kind of work where the question is not just "what happened" but "why did the metric change, which segment drove it, is the change statistically meaningful, and what should the team do next?"

Built with Python, SQL, and dbt.

Data & Analytics: SQL · Python · pandas · NumPy · Jupyter

Analytics Engineering: PostgreSQL · dbt · data modeling · ETL pipelines · data quality checks

BI & Visualization: Power BI · Tableau · Streamlit · Plotly

ML & Statistics: scikit-learn · XGBoost · SHAP · statistical testing · model evaluation

Governance: data quality rules · metadata layers · audit logs · policy checks · dataset health scoring · AI governance patterns · agent governance

Tools: Git · GitHub · Docker · pytest

I'm not going to list fifteen bullet points that all say "I'm great at everything." Here is what I consistently do:

• Turn raw data into structured, tested analytics layers — not fragile one-off queries that break when someone changes a column name.
• Build dashboards backed by reliable data models where the metric definitions are documented and the data quality is checked before anything reaches a stakeholder.
• Connect analytics work to the actual business question. The dashboard is not the deliverable. The decision it supports is.
• Think about governance at both the dataset level and the AI-agent level : who has access, what rules apply, what evidence exists.
• Explain technical systems to non-technical people without dumbing things down or hiding the complexity that matters.

• LinkedIn: linkedin.com/in/someshzanwar
• Portfolio: someshzanwar.github.io
• Email: someshzanwar345@gmail.com