Pinned Loading

1. llm-eval-harness llm-eval-harness Public

   Automated RAG evaluation framework: LLM-as-judge scoring for retrieval quality, faithfulness (hallucination), and correctness — 100% faithfulness, cost/latency tracked per query. Benchmarks hybrid …

   Python

2. pipeline-doctor pipeline-doctor Public

   Agentic data-remediation system: a bounded Claude tool-use agent fixes BigQuery anomalies (nulls, zeros, outliers) with pandas in a sandbox, verifying each fix by re-running detection with regressi…

   Python

3. curriculum-engine curriculum-engine Public

   Hybrid RAG pipeline: Claude Haiku plans a curriculum via search queries; YouTube/Serper APIs retrieve and validate live links. Prompt caching cuts cost up to 90%. Standalone benchmarks: FAISS index…

   Python

4. customer-event-pipeline customer-event-pipeline Public

   3-stage Kubeflow Pipelines (KFP v2) ML pipeline on Vertex AI: BigQuery extraction → feature engineering → scikit-learn model training with full artifact lineage tracking. Components run in isolated…

   Python

5. customer-event-flink customer-event-flink Public

   PyFlink streaming pipeline: Kafka → 5-minute tumbling event-time windows → per-customer spend aggregation. Handles late-arriving events with watermarks and BoundedOutOfOrderness strategy.

   Python

6. snowflake-pipeline snowflake-pipeline Public

   End-to-end ELT pipeline: Python → Snowflake → dbt (3-layer modeling) → Apache Airflow orchestration. Incremental loading, data quality tests, GitHub Actions CI.

   Python