Marketplace

Use when a retrieval corpus has exact or near-duplicate docs/chunks inflating embedding cost or polluting top-k results, and you must detect and prune them.

Invoke when building or tuning a RAG/semantic-search pipeline's chunking: pick strategy, chunk size, and overlap validated against retrieval metrics.

Use when auditing an LLM answer that carries source citations, to verify each claim is entailed by its cited sources and flag unsupported or misattributed ones.

Design conversational analytics interfaces: NL-to-SQL, query disambiguation, result narration, and data storytelling.

Design ML experiment tracking: hyperparameter logging, model versioning, reproducibility, and experiment comparison workflows.

Design and implement ML feature stores: online/offline serving, feature freshness, training-serving skew prevention, and feature discovery.

Choose the right chart form for a dataset and question, then build an honest, labeled, accessible visualization; use whenever data needs plotting.

Plan ML model deployment: serving patterns, A/B testing, canary releases, monitoring, and rollback for production ML.

Build comprehensive model evaluation: benchmark selection, statistical significance, human evaluation protocols, and safety testing.

Design recommendation systems: collaborative filtering, content-based, hybrid approaches, cold start, and ranking.

Design data governance frameworks: data classification, lineage tracking, access policies, retention, and compliance mapping.

Curate and clean training datasets: deduplication, quality scoring, bias detection, and data augmentation strategies.

Systematically clean a messy dataset with profiling, ordered fixes, validation, and a provenance log; use before analyzing or merging untrusted data.

Design robust data pipelines: ingestion patterns, transformation stages, schema evolution, and data quality monitoring.

Design data quality monitoring: anomaly detection, schema validation, freshness checks, and data contracts between teams.

Design data warehouse schemas: star/snowflake schemas, slowly changing dimensions, materialized views, and query optimization.

Design document processing pipelines: OCR, layout analysis, information extraction, classification, and human review.

Choose embedding models and strategies: dimensionality, domain-specific models, multi-modal embeddings, and evaluation.

Design and scaffold idempotent, observable extract-transform-load pipelines with safe retries, backfills, and quality gates; use when building data pipelines.

Optimize model inference: quantization, distillation, batching, speculative decoding, and serving infrastructure tuning.

Plan and execute LLM fine-tuning: dataset preparation, method selection (LoRA, QLoRA, full), evaluation, and deployment.

Plan and execute safe database schema migrations: dependency ordering, rollback strategies, zero-downtime patterns, and data backfill procedures.

Diagnose slow SQL with execution-plan analysis and rewrite queries and indexes; use when a query is slow, times out, or burns database CPU.

Design real-time stream processing: windowing, watermarking, exactly-once semantics, and state management.