3. Data, Science & AI

Automates extraction and upload of Dacon competition datasets to Hugging Face Hub for AI/ML model development and sharing.

Validates ClickHouse and Kafka data pipelines with defense-in-depth patterns, including producer/consumer validation, error streaming, and schema evolution.

Assists in creating and editing Excel spreadsheets with formulas, financial modeling, and data analysis standards.

Captures and analyzes the current game state from a screenshot, enabling game state interpretation through image processing or AI.

AI-powered research tool for analyzing GitHub, GitLab, and Bitbucket repositories to understand codebases and project architecture using RAG.

Enterprise RAG solution for production-ready document retrieval, semantic search, and organizational knowledge management via CustomGPT.ai.

Creates and optimizes AGENTS.md files to define AI coding agent instructions, ensuring adherence to project conventions and improving agent performance.

Provides best practices for implementing structured logging in data collection pipelines to ensure consistent and actionable log data.

Ensures AI responses are truthful and evidence-based, especially during user requests for honesty or high-error-cost scenarios.

Orchestrates sequential execution of specialized AI agents to complete complex development, security, and maintenance tasks through coordinated workflow management.

Context management system for LLM applications, providing TLDR summaries, conversation tracking, and post-compaction recovery via '/baton' trigger.

Validates AI agents against production quality criteria across nine dimensions, delivering a detailed score and actionable improvement report.

Automates Excel spreadsheet creation, editing, and analysis using Python libraries, supporting formulas, formatting, and financial modeling with industry-standard conventions.

Creates the LearnFlow database schema on Neon PostgreSQL, including table structures and relationships.

Calculates portfolio risk metrics including VaR, CVaR, Sharpe, Sortino, and drawdown for risk assessment and monitoring.

Generates speech audio and clones voices locally using open-source AI tools like Coqui TTS and Bark, avoiding paid APIs.

Augments existing ideas with contextual data from OpenContext, enhancing concept depth through knowledge base integration.

Provides best practices for building efficient and maintainable web scraping pipelines in Python, including handling common challenges and data extraction patterns.

Builds AI agents with OpenAI Agents SDK, supporting async patterns, multi-agent orchestration, and tool-calling for tutoring and specialist workflows.

Creates specialized LLM skills via template-driven generation, enabling rapid development of AI capabilities.

Optimizes AI model context by compressing input data to reduce token usage and memory footprint.

Assists in evaluating AI model performance using pass@k, regression testing, and quality metrics.

Fine-tunes, evaluates, and deploys Helsinki-NLP OPUS-MT models for Chuukese-English translation using custom data.

Generates new AI agent skills following the Agent Skills specification for custom LLM behavior definition.