APPLIED AI & DATA SYSTEMS

From a messy problem
to a useful system.

I’m Tong. I build the models, data foundations, and feedback loops that help AI work in real operations.

My work connects prediction and evaluation with the engineering that keeps a system trustworthy after it ships.

Explore my work

SELECTED WORK / NUVOCARGO

Different problems.
A connected practice.

Logistics gave me a practical test for every model: can someone use this to make a better decision?

01 / PREDICTION

Appointment prediction

Turn scheduling history into predictions an operational workflow can safely act on.

PythonEvaluationSnowflake

Explore the work

The problem

Appointment automation needs more than a plausible date. It needs reliable ground truth, customer context, and a clear boundary for when to act.

My contribution

Improved prediction logic, built reproducible replay evaluation and an experiment loop, and contributed storage, service integration, rollout safeguards, and funnel measurement.

What I learned

Offline improvement and accuracy within a selected customer cohort answer different questions. Keeping those populations distinct makes rollout decisions more honest.

APPLIED AI

02 / INTELLIGENCE

Carrier & freight intelligence

Connect external signals with operational data to support procurement and pricing decisions.

dbtData integrationDecision modeling

Explore the work

The problem

Vendor feeds have different identifiers, coverage, update patterns, and definitions. A useful market signal starts with resolving those differences.

My contribution

Built ingestion and warehouse models, investigated reconciliation and scoring parity, and developed carrier-presence and freight-market analyses.

What I learned

A research improvement becomes business value only when a policy is adopted and its operational effect is measured. Those are separate milestones.

DATA & MODELING

03 / FOUNDATIONS

Data systems people can trust

Make warehouse data accessible to humans and AI, with meaning, boundaries, and feedback.

MCPSemantic contractsObservability

Explore the work

The problem

A working query is not necessarily the right answer. Analytics needs consistent definitions, controlled access, and traceable execution.

My contribution

Designed semantic data-access contracts and contributed catalog, query, retrieval, and evaluation work. Built and maintained internal BI, schema-sync automation, and shared engineering workflows.

What I learned

Reliability includes the people building alongside you. Reusable deployment and review practices help good decisions travel beyond one repository.

PLATFORM

Selected contributions from my work at Nuvocargo. These descriptions focus on my role within collaborative projects.

HOW I WORK

Evidence before
confidence.

I like problems where the answer has to survive contact with the operation.

Start with the decision.

Understand the customer, workflow, constraints, and cost of being wrong before choosing a model.

Make improvement testable.

Build ground truth and repeatable evaluations. Investigate unexpected results rather than hiding them in an average.

Own the path after launch.

Bring the data, deployment, monitoring, and review practices along with the algorithm.

A LITTLE CONTEXT

Builder.
Modeler.
Systems thinker.

My Nuvocargo work spans operational analytics, route optimization, document automation, prediction, and data platforms.

The connecting thread is translation: from a customer’s constraint to a model, from a model to a service, and from a service to something the team can maintain.

I’ve also worked directly with customers on route recommendations and helped colleagues adopt deployment and engineering practices. I’m interested in teams building useful AI products and the data systems behind them.

2024

Operational analytics & optimization

2025

Document automation & agent evaluation

2026

Prediction & production data/AI systems

LET’S TALK

What are you building?

I’d like to meet people working on applied AI, reliable data products, and difficult operational problems.

Email Tong GitHub

From a messy problemto a useful system.

Different problems.A connected practice.

Appointment prediction

The problem

My contribution

What I learned

Carrier & freight intelligence

The problem

My contribution

What I learned

Data systems people can trust

The problem

My contribution

What I learned

Evidence beforeconfidence.

Start with the decision.

Make improvement testable.

Own the path after launch.

Builder.Modeler.Systems thinker.

What are you building?

From a messy problem
to a useful system.

Different problems.
A connected practice.

Evidence before
confidence.

Builder.
Modeler.
Systems thinker.