3. Data, Science & AI

Expertly constructs and optimizes Datalog queries for Logseq's graph database, covering syntax, performance, and advanced patterns including pull and aggregations.

Provides authoritative knowledge of Logseq's Datascript database schema, including attributes, relationships, and node/block data model architecture.

Expertly creates, modifies, and validates Claude Code agents and their YAML configurations, including architecture design and component implementation.

Creates reusable, themeable plot recipes in Makie for custom data visualizations in Julia.

Comprehensive technical documentation and best practices for DolphinDB, a high-performance time-series database, including design, stream computing, and quantitative backtesting.

Optimizes CUDA kernels and GPU code for peak performance in high-performance computing and AI applications.

Advises on optimizing code for NVIDIA GPU architectures (Ampere, Hopper, Ada) to improve performance in compute-intensive applications.

Automates web data extraction from websites, including JavaScript-heavy pages, with structured output and schema generation for efficient data pipelines.

Migrates prompts and code from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, adjusting for model behavior differences.

Converts SQL migration scripts to OSC format for Flyway, enabling online schema changes without table locks during database migrations.

Optimizes AI-assisted development through token budget management, context compression, and degradation detection using context-engineering principles.

Deeply analyzes LLM agent and plugin repositories to extract architecture, patterns, and documentation insights.

Facilitates structured agent-to-agent messaging using fmail, adhering to team communication conventions for collaborative AI workflows.

Conducts statistical analysis, data visualization, and generates publication-quality reports from datasets using Python.

Safely queries internal data sources to generate reproducible, auditable analytics artifacts for internal reporting and decision-making.

Manages LLM prompt templates using Mustache for dynamic substitution, enabling reusable and structured prompt generation in AI applications.

Automated evaluation system for RAG pipelines and LLM agents, using LLM-as-judge for quality assessment and retrieval analysis.

Vector search library for semantic search and RAG, enabling metadata-filtered, threshold-based retrieval of embeddings.

Optimizes LLM context windows by constructing token-limited conversation history and streamlining prompt building for efficient language model interactions.

Automates document processing by converting PDFs, PPTs, and images into Schema.org annotated HTML through configurable transformation steps.

Converts PDF documents to HTML using LLM vision capabilities, splitting pages to images and processing with multimodal models for document conversion and knowledge extraction.

Provides an RDF graph index for storing and querying linked data triples using the N3 library, supporting semantic queries and SHACL constraints for knowledge graphs.

Orchestrates LLM-based conversational agents with RAG, memory management, and tool execution for multi-turn dialogues.

Unified LLM API client for OpenAI-compatible endpoints, handling chat, embeddings, and streaming.