3. Data, Science & AI

Creates parameterized prompt templates for AI development, encoding domain patterns to reduce repetition and improve efficiency.

Optimizes internal memory state of AI agents to improve context retention and response accuracy during interactions.

Enables hypothesis testing, A/B testing, and regression analysis for data-driven decision-making.

Enables effective data visualization through chart selection, dashboard design, and storytelling using Tableau and Power BI.

Enables semantic code search and codebase understanding through AI-powered retrieval using code embeddings and knowledge graphs.

Enables advanced data analysis in Excel with pivot tables, formulas, and Power Query for efficient data transformation and reporting.

Provides token-efficient content retrieval from files using hierarchical depth levels (L0-L5) and heading-based sectioning for structured data access.

Enables advanced SQL data analysis through complex queries, joins, aggregations, window functions, and query optimization.

Extracts and explains mathematical formulas from academic texts in Physical AI and Humanoid Robotics, aiding conceptual understanding.

Inspects Supabase database schemas and data for read-only queries, such as checking tables, listing columns, or retrieving data.

Tool for deduplicating machine learning datasets using MinHash, LSH, and clustering to enhance data quality and reduce redundancy.

Enables GPU scheduling with VRAM budgeting, work queues, and model lifecycle management for ML pipeline orchestrators.

Enables efficient checkpoint compression for PyTorch models using byte grouping and optimized I/O pipelines to reduce storage overhead.

Enables GPU-accelerated machine learning inference in Rust via the Candle framework, eliminating Python dependencies for ML pipelines.

Enables building GPU-accelerated machine learning models in Rust using the Candle framework for vision, LLM, and audio applications.

Reads and converts Apache Avro files to readable JSON, automatically deserializing nested fields for clarity.

Port HuggingFace models to Candle framework, resolving tensor naming, weight loading, and matmul issues during migration.

Provides accurate, evidence-based answers by referencing documents, textbooks, and research papers.

Designs AI-driven NPC dialogue systems for escape rooms using adaptive responses, emotional simulation, and player profiling with Gemini/GPT-4.

Analyzes ICU clinical data in CLIF format, computes SOFA, CCI, and Elixhauser scores, and transforms datasets for research.

Transcribes Vaam video content into text using Google Gemini AI, supporting any language without translation. Accepts a share URL and returns full transcription.

Provides a command-line interface for Google's Gemini AI model to process images, PDFs, and large files with 1M+ token context and domain-specific reasoning.

Enables consistent capture of preferences, procedures, and factual data through Graphiti MCP tool workflows.

Self-improving context loader that retrieves relevant past project data from Graphiti to inform and enhance prompts with tailored recommendations.