Evaluation pipelines, data platforms, and tools that help make complex systems easier to understand.
Offline Evaluation
Built an offline evaluation pipeline measuring retrieval quality, hallucination risk, latency, and cost.
Produces deterministic outputs that can be checked in CI to make changes in retrieval and answer quality easier to inspect.
View repositoryEnterprise RAG Simulator
Built an enterprise RAG platform simulator with hybrid retrieval and access-aware answers.
Includes audit logs and an operator dashboard for exploring governance and observability workflows.
View repositoryRelease-Gate Simulator
Built a release-gate simulator evaluating pipeline quality, safety, latency, and cost.
Produces auditable PASS, WARN, and BLOCK decisions to make release criteria and their outcomes inspectable.
View repositoryRepository Intelligence
Open-sourced a local-first repository-intelligence server that indexes codebases into dependency graphs.
Supports code and symbol search, architecture exploration, and change-impact analysis across a repository.
View repositoryLocal agent skill and research engine that compares AI, data, and developer tools using sourced adoption signals, scoring, and generated briefs.
Built for decision-ready tool evaluation with reproducible local artifacts and inspectable engineering structure.
Deterministic multi-agent testbed for reliability, safety, observability, prompt-injection defense, retries, budget guards, and scorecard evaluation.
Focuses on repeatable agent failure modes and practical release-readiness checks.
Deterministic local workflow engine for planner, executor, and critic agent flows with retries, budgets, checkpoints, traces, and replay.
Shows practical agent orchestration mechanics beyond a single prompt-response loop.
Offline real-time decisioning simulator for fraud, risk, and marketing events with streaming features, explainable scoring, drift monitoring, and dashboards.
Combines feature generation, scoring, monitoring, and business-facing analytics in one local system.
Data reliability control plane that validates dataset contracts, tracks lineage blast radius, evaluates freshness SLAs, and generates incident dashboards.
Maps upstream data quality failures to downstream business impact and operational response.
Local marketing analytics lakehouse simulator with bronze, silver, and gold pipelines, schema validation, KPI marts, quality gates, and a static dashboard.
Reflects practical analytics engineering patterns from raw ingestion through modeled reporting outputs.
Python time-series analysis project for stock-market data, covering trend detection, stationarity testing, differencing, moving averages, and visualization.
Shows statistical analysis workflow from exploratory plots through stationarity-focused transformations.
Machine learning analysis project for credit-card approval decisions, covering data cleaning, feature preparation, model evaluation, and approval-risk insights.
Demonstrates a structured classification workflow for tabular decision data.
Math generation project focused on creating structured math content and exercises with repeatable generation, simple workflows, and inspectable outputs.
Applies automation patterns to educational content generation.