3. Data, Science & AI

Enables integration of large language models (LLMs) for structured data extraction, prompt engineering, and multi-step AI workflow implementation.

Generates scientific experimental procedures from LA-Bench JSONL data through AI subagent orchestration for parsing, validation, and output formatting.

Generates step-by-step experimental procedures with quantitative specifications from LA-Bench input, using delegated logic for proper experimental design.

Expert in designing systematic AI prompts using Socratic questioning to enhance information extraction accuracy and reliability.

Refines LA-Bench experimental procedures through iterative validation and regeneration cycles, improving quality via a loop between checker and generator skills.

Extracts structured experimental protocol data from LA-Bench JSONL files by experiment ID, including materials, steps, outcomes, and references.

Intelligently scrapes RBI bulletins to extract and structure information about Indian government securities for financial analysis.

Computes text embeddings and measures semantic similarity between sentences for natural language processing tasks.

Provides a precomputed musical taste profile for Laimis, enabling AI music prompt generation, mood-based recommendations, and style analysis without reprocessing data.

Normalizes metadata for excavation investigation materials including papers, reports, and adjacent relics to ensure consistent data structure.

Generates text-based visualizations and reports showing the relationship between spelling errors and semantic distance in experimental data.

Updates genealogy database records for birth/death dates, marriages, children, and biographical data while ensuring relationship consistency across 10 lineages.

Builds functional data pipelines on Keboola for extraction, transformation, validation, and dashboard creation.

Toolkit for managing Hugging Face Spaces, including listing, space information retrieval, runtime control (restart/pause), and status checks for AI model deployments.

Provides RDBMS access patterns for DuckDB, MySQL, PostgreSQL, SQL Server, and DBISAM using ODBC and native drivers.

Validates day-of-week accuracy, relative dates, and countdowns in planning and calendar data using dual-source verification.

Automates QuantConnect cloud backtesting workflows including strategy deployment, execution, and performance metric retrieval via GitHub integration.

Extracts structured JSON data from business document images (invoices, receipts, etc.) using Vision API for automated digitization.

Automates audio transcription and image OCR for Gastrohem WhatsApp media folders using AI models, generating structured JSON and markdown summaries.

Analyzes Snowflake query performance, identifies optimization opportunities, and provides cost reduction recommendations for efficient data operations.

Generates and manages dbt models for data transformation, ensuring best practices in SQL, configuration, and documentation for staging, intermediate, and mart layers.

Analyzes dbt model dependencies, traces data flow, identifies circular dependencies, and visualizes data lineage for impact analysis and source tracing.

Provides deterministic orchestration for multi-agent LLM systems in LMStudio, enabling constraint-driven agent composition and resource allocation.

Analyzes KPI metrics from Excel data and generates structured markdown reports based on assigned task divisions.