3. Data, Science & AI

Comprehensive documentation and tools for single-cell RNA-seq analysis, including API, preprocessing, analysis, and visualization with Scanpy.

Analyzes Excel/CSV fields, auto-generates Chinese-English mappings, validates translations, and outputs statistical reports for data dictionaries.

Comprehensive reference atlas framework for single-cell deep learning with full API documentation, tutorials, and multimodal integration capabilities.

Generates structured summaries of text documents, optionally archives to topic profiles, and stores in U1's long-term memory system.

Provides documentation and utilities for PostgreSQL 16 database operations including SQL, JSON, full-text search, and performance tuning.

Provides vector similarity search capabilities in PostgreSQL, supporting embeddings storage and advanced indexing for AI/ML applications.

Orchestrates multi-agent workflows for the Pomodoro Time Tracker, enabling handling of complex tasks and test failures.

Optimizes machine learning training performance in Kaggle competitions by resolving slow training and GPU memory (OOM) issues through parameter adjustments.

Refines vague user prompts into structured, high-context instructions for AI coding agents using Chain-of-Thought reasoning, optimizing interactions with models like Codex and Claude Code.

Integrates Trackio for logging metrics and managing experiments during model training in Kaggle competitions.

Integrates AI capabilities into Directus for chat interfaces, content generation, and smart suggestions, enabling AI-powered content management.

Python library for extracting text, tables, images, and parsing PDF documents efficiently.

Optimizes system prompts for Claude Code agents using proven prompt engineering techniques to improve agent performance and workflow effectiveness.

Analyzes training patterns, performance, and physiological data using computational models to provide insights on load, recovery, and readiness.

Guides users through complete Kaggle competition workflow from experiment setup to submission preparation.

Creates parameterized prompt templates for AI development tasks, enabling reusable patterns and reducing cognitive load in AI interactions.

Builds a FastAPI-based RAG service for textbook content using OpenAI, Neon Postgres, and Qdrant for ingestion pipelines and query endpoints.

Provides personalized concert and live music recommendations based on user preferences and music taste for bucket list experiences and upcoming shows.

Creates custom AI agents for task automation and interaction using natural language processing and machine learning.

Formats structured data (tables, logs, API responses) into TOON v2.0 for consistency in RAG pipelines and data processing workflows.

Analyzes order data to determine profitable free shipping thresholds using order value distribution, hero product analysis, and shipping economics.

Performs quantitative analysis, factor diagnostics, and reporting on ATFT-GAT-FAN model outputs for data-driven insights.

Manages J-Quants data ingestion, feature graph generation, and cache hygiene for the ATFT-GAT-FAN dataset pipeline.

AI-powered memory system that tracks project interactions, learns from corrections, and maintains token-efficient context across sessions to preserve institutional knowledge.