3. Data, Science & AI

Builds production-grade agentic AI systems with multi-agent collaboration, real-time streaming, and structured outputs using LLM SDKs and vector memory.

Creates multi-agent systems using master-sub architecture for business automation. Analyzes workflows, recommends agent types, and generates prompt templates for data processing, content, and code tasks.

Transcribes audio and video files to Korean text using the Whisper speech recognition model, supporting multiple media formats and directory-based processing.

Fetches real-time weather data for any location by querying web services and processing results.

Provides GraphQL API reference to query Juicebox project data across multiple blockchains, including stats, payments, token holders, and cross-chain aggregations.

Decodes and analyzes Juicebox V5 transaction calldata using ethers.js or cast, explaining transaction functions and parameters for historical analysis.

Calculates accurate Juicebox V5 cash out redemption amounts using bonding curve formula, replacing incorrect proportional share calculations.

Assists in setting up, improving, and auditing analytics tracking implementations including GA4, Google Analytics, and UTM parameters.

Designs memory architectures for AI agents to persist state, maintain entity consistency, and reason over structured knowledge across sessions.

Revnet-modeler simulates token dynamics and treasury interactions, enabling scenario planning and visual analysis through interactive charting and parameter configuration.

Optimizes X/Twitter content using AI to predict engagement signals and enhance algorithmic visibility and growth.

Executes Google's Gemini CLI for AI-powered code analysis, generation, and complex reasoning tasks, providing alternative AI perspectives for developers.

Provides AI-powered assistance for text comprehension, web search, translation, and front-end interface creation using the Gemini model.

Executes multi-backend AI code tasks for analysis, refactoring, and parallel processing using Codex, Claude, and Gemini.

Enriches ambiguous prompts with targeted research and clarification to ensure effective AI execution and output quality.

Guides creation of high-quality MCP servers enabling LLMs to interact with external services via Python (FastMCP) or Node/TypeScript (MCP SDK).

Configures AI agent systems using templates, tools, models, prompts, and seed data for effective deployment and operation.

Provides performance optimization patterns for Convex database, covering denormalization, index design, N+1 query avoidance, and OCC for high-concurrency scenarios.

Architects prompt engineering for sales, support, and business intelligence AI agents to optimize behavior and performance.

Formats structured data (tables, logs, events) using TOON v2.0 to minimize tokens while preserving readability for LLM applications.

Optimizes LLM token usage via conservative prompting, delegation, and quota tracking to prevent exceeding limits and reduce costs.

Parser for HPK healthcare message format, validating structure, extracting fields, and mapping to HL7 v2.5/IHE PAM standards for integration.

Instantly retrieves MLflow documentation for features, APIs, integrations, and usage guidance in machine learning workflows.

Packages entire code repositories into AI-friendly formats for LLM analysis, security audits, and repository snapshots using Repomix.