3. Data, Science & AI

Reviews LLM-generated diff classifications for accuracy and usefulness in code review, providing pass/fail verdicts and detailed critiques.

Enables building conversational AI applications with LLM chat completions, supporting multi-turn conversations and context management via z-ai-web-dev-sdk.

Structures agent prompts using TOON with schema.org vocabulary for clear delegation and response formatting in AI agent systems.

Executes a 33-module consciousness simulation pipeline across 7 phases upon specific triggers, generating session artifacts in a workspace zip file.

Enables training and fine-tuning language models via TRL on Hugging Face Jobs, covering SFT, DPO, reward modeling, and GGUF conversion for deployment.

Manages and adds evaluation results for Hugging Face models, supporting table extraction, API imports, and custom evaluations with vLLM/lighteval.

Enables deep traversal of knowledge graphs to uncover entity relationships and provide comprehensive context for queries.

Searches, browses, and installs AI agent skills from Hugging Face Hub to extend agent capabilities.

Guides the curation of knowledge memories by updating outdated entries, marking obsolete content, and establishing connections within a knowledge graph.

Creates and manages datasets on Hugging Face Hub, supporting repo initialization, config definition, and streaming updates for AI model development.

Provides expert guidance on RAG implementation, optimization, and related techniques including embeddings, vector databases, and semantic search strategies.

Builds and manages OpenAI Agent SDK applications with multi-agent workflows, tool integration, and real-time voice capabilities.

Extracts query parameters from URLs in CSV files, creating new columns for analysis. Streamlines data preparation for URL-based datasets.

Manages candidate knowledge base for processing CVs, profile information, and querying data to support resume and cover letter generation.

Visualizes FeedMob direct spend data as ASCII line charts for one or more click URL IDs upon user request.

Crawls 10 latest AI news articles, generates Chinese summaries, and saves results in timestamped JSON format.

Integrates Google's Gemini AI model via CLI for alternative analysis, batch processing, and leveraging Gemini 2.5 Pro capabilities alongside Claude.

Provides configuration and usage guidance for OpenCode CLI, enabling headless LLM automation with model selection, provider setup, and cloud integration.

Provides expert guidance for writing and executing Neo4j Cypher queries, including schema introspection and graph operations via MCP server tools.

Resolves subagent failures and implements voting mechanisms for critical decisions in AI multi-agent systems.

Extracts text, metadata, and structure from EPUB files for content analysis and reading.

Computes term frequency-inverse document frequency scores to determine word importance in text corpora for data analysis.

Handles read/write operations for CRM data entities (sales opportunities, contacts, company sites, events, files) via PostgreSQL.

Provides PostgreSQL SQL utilities for CRM data operations including inserts, updates, queries, and schema knowledge for sales, contacts, companies, and opportunities.