3. Data, Science & AI

Analyzes meeting transcripts to identify communication patterns and provide actionable feedback for improving leadership and interpersonal skills.

Creates architecture and system diagrams from codebase analysis in Excalidraw format for visualizing code structure and system design.

Downloads and converts Dutch legal documents from official government databases into structured YAML format for data integration.

Analyzes Mixpanel event data, user profiles, and analytics using SQL, JQL, and DuckDB for data processing and insights.

Queries UniProt, STRING, and BioGRID for protein interactions, functional enrichment, and ID mapping.

Validates synthetic lethality claims from CRISPR knockout screens using BioGRID ORCS and essentiality scoring methods (CERES, MAGeCK, BAGEL).

Constructs life sciences knowledge graphs by orchestrating APIs and applying the Fuzzy-to-Fact protocol to map biological connections and drug pathways.

Queries ChEMBL, PubChem, and DrugBank for drug mechanisms, targets, and bioactivity data.

Queries Open Targets and ClinicalTrials.gov for target-disease associations, drug target validation, and clinical trial discovery.

Queries genomic databases (Ensembl, NCBI, HGNC) for gene annotations, variant analysis, and ID resolution using scientific identifiers.

Analyzes and explains X algorithm engagement scores to determine post ranking factors, score weights, and engagement potential.

Explains the X recommendation algorithm pipeline, detailing how posts are ranked and the system's underlying mechanics.

Explains Phoenix ML model architecture for X recommendations, detailing embeddings, transformers, and prediction mechanisms.

Checks remaining quota, usage limits, and rate limits for AI services including Claude Code, OpenAI Codex, Google Gemini, and Z.AI via CLI.

Executes user-defined prompts via various AI models from a designated directory, supporting task delegation and verification loops in terminal environments.

Enables distributed coordination between AI agents for shared project management, task handling, and institutional knowledge access within the HACS network.

Assists in database schema design, query optimization, and performance tuning for both SQL and NoSQL databases.

Analyzes session history to extract and store learnings, ensuring the system remembers corrections and applies them in future interactions to avoid repeating mistakes.

Integrates Obsidian vaults with Claude Code for AI-driven knowledge management, enabling auto-linking, bidirectional sync, and self-evolving knowledge graphs.

Packs codebases into structured, AI-optimized formats for efficient LLM analysis, enabling code review and summary generation.

Converts PDF, Word, web content, and structured data to Markdown for data pipelines and LLM preprocessing tasks.

Prototype tool for extracting and preparing training data specifically for Qwen, a large language model.

Validates weather station data sources, checks live condition coverage, and identifies locations with missing real-time weather data.

Prototype tool for fine-tuning Google's Gemma language model on domain-specific data.