Foundations and models
Explain transformers, tokenization, inference, and model selection through explicit trade-offs.
An executable comparison and a model decision backed by evaluation.
YOUR WORK · IN YOUR BROWSER
A progress bar says what you opened. A portfolio proves what you designed, measured, and operated. Use this space to turn every competency into a reviewable artifact.
Expected evidence
Status is a local self-assessment. “Verified” means you retain an artifact, a metric, and a review; it is not an official credential.
Explain transformers, tokenization, inference, and model selection through explicit trade-offs.
An executable comparison and a model decision backed by evaluation.
Treat prompts, schemas, and tools as versioned, testable contracts.
A regression dataset and A/B pipeline with explicit criteria.
Design ingestion, chunking, search, reranking, grounding, and RAG evaluation.
A measured RAG system with answerable and unanswerable cases, citations, and retrieval metrics.
Build graphs, tools, memory, and failure recovery without losing control.
An agent with three tools, traces, idempotency, and negative-path tests.
Separate model, retrieval, and application quality through reproducible evaluations.
A regression suite with a baseline, rubric, failed samples, and acceptance gates.
Observe latency, cost, quality, safety, and distribution shifts in production.
An operational dashboard, alerts, and an incident runbook.
Model threats, constrain tools, and apply privacy, guardrails, and auditability.
A threat model, prompt-injection test, and verifiable controls.
Make build-vs-buy, cost, reliability, and user-experience decisions.
ADRs, cost analysis, and a technical capstone defense.
Download JSON to move progress between devices or keep a copy. Import replaces local state only after validation.