3. Data, Science & AI

Measures machine learning model performance on test datasets using accuracy, precision, and recall metrics.

Extracts and interprets mathematical formulas from research papers for direct implementation in data science or AI models.

Applies SIMD optimizations to Mojo code for accelerating tensor and array operations in AI and high-performance computing.

Reads and integrates responses from Codex, enabling AI-assisted code generation and integration in development workflows.

Adapts SQL for TiDB, resolving differences from MySQL including VECTOR types, full-text search, and transaction modes for compatibility and migration.

Provides Python utilities for TiDB database interaction with AI features: vector search, embeddings, RAG, and text2sql examples.

Sends code to Codex AI for review via tmux file-based IPC, triggered by /codex-review command.

Integrates LLM providers with Pydantic for model configuration, fallback strategies, and streaming API handling to optimize AI model usage.

Enables vector similarity search in SQLite for embedding storage, KNN queries, and semantic search features using the sqlite-vec extension.

Guides architectural decisions for LangGraph applications, including state management, multi-agent system design, and persistence strategies.

Creates type-safe AI agents with Pydantic validation for structured LLM outputs and seamless integration in chat systems.

Document parser for 15+ formats (PDF, DOCX, PPTX, HTML, images) enabling text extraction, Markdown/HTML/JSON conversion, and RAG chunking.

Enables AI agent implementation via Deep Agents framework, including agent creation, backend configuration, subagent definition, and middleware integration.

Offers UI components for AI workflows including chat interfaces, tool execution displays, and reasoning visualization.

Implements stateful agent graphs with LangGraph for building multi-agent AI systems, including state management and checkpointing.

Analyzes accumulated feedback data to identify patterns and provide actionable suggestions for improving review accuracy and effectiveness.

Researches and develops semantic theories using ModelChecker with Z3 SMT solver, including operator definition, constraint adjustment, and testing.

Analyzes transcripts to extract speaking style, phrases, and vocabulary patterns for authentic author profile creation.

Maps LinkML enum permissible values to ontology terms, resolving CURIE validation errors and ensuring accurate semantic data representation.

Guides selection of specialized AI agents for specific task types to optimize system performance and resource allocation.

Enables creation, editing, and analysis of spreadsheets with formula support, formatting, and data visualization for structured data handling.

Diagnoses AI skill prompts for token waste, anti-patterns, and trigger issues, delivering actionable optimization suggestions for improved performance and publication readiness.

Ensures formula, formatting, and multi-sheet relationships remain intact when modifying existing Excel workbooks.

Comprehensive library of proven prompting patterns, frameworks, and examples for effective AI model interaction and prompt engineering.