3. Data, Science & AI

Converts session episodes into persistent memory patterns for enhanced AI context retention and personalization.

Provides reference for correctly spawning subagents to delegate specialized tasks within an AI agent framework.

Guides parsing of Cassandra 5.0+ SSTable files, including compression handling and structure analysis for debugging and data inspection.

Guide for designing, evaluating, and optimizing AI prompt templates using techniques like few-shot and chain-of-thought.

Creates .prompt files for LLM workflows in Output SDK, enabling prompt design, LLM provider configuration, and Liquid.js templating.

Analyzes prompts within Flow SDK workflows to identify formats, variables, and conversion requirements for AI model preparation.

Builds scalable data pipelines in Elixir using GenStage, Broadway, and Flow for stream processing.

Generates D&D character images using AI models (Gemini or Whisk) via browser automation for manual interaction.

Assists in designing comprehensive test cases for LLM systems using assay-kit, addressing hallucinations, overfitting, and partial processing failures.

Analyzes AI agent execution logs to compute and report time metrics for file preparation, retrieval, code generation, and verification stages.

Analyzes PDFs, images, video, audio, charts, and screenshots to extract data, describe visuals, and explain diagrams using AI.

Creates RAG-ready component cards from system-map/components.yaml, structuring component details to replace full code with essential context for LLMs.

Archives conversation transcripts with AI embeddings and analysis in ArangoDB, preserving user stories and chain of thought for retrieval.

Maintains PCE memory integrity by resolving errors, duplicates, and outdated claims using pce_memory_feedback.

Enables complex data analysis and querying across CRM records using SQL-like syntax for filtering, grouping, and aggregation.

Standalone FastAPI server for millisecond-latency embeddings, supporting model swapping and integration with vector databases like Pinecone and ArangoDB.

Extracts structured content from diverse document formats including PDF, DOCX, and images using automated format detection and agentic pipelines.

Manages scillm Paved Path execution for batch text, VLM, and Lean4 proving with parallel processing capabilities.

Collects and verifies project context for knowledge base construction using AI validation. Designed for onboarding, codebase analysis, and knowledge management.

Transient vector store service leveraging FAISS for high-speed similarity search, optimized for AI/ML applications like semantic search and recommendation systems.

Toolbox for AI-native development leveraging MCP servers to enable context-aware automation and seamless AI model integrations.

Fetches and extracts content from web pages, PDFs, and documents with automatic fallbacks for reliable data retrieval.

Builds ML recommendation systems using collaborative, content-based, and hybrid models with solutions for cold-start and diversity.

Analyzes Excel file structure to identify worksheets, headers, formulas, and data types for data processing and integration.