3. Data, Science & AI

Executes reproducible Python-based data analysis, statistical testing, visualization, and predictive modeling for dataset exploration and insights.

Expert assistance for implementing Google's GenAI SDK v1.35.0 with structured output control and response formatting capabilities.

Provides a base instruction protocol for developing and managing advanced AI agentic systems in senior engineering contexts.

Runs Microsoft's eval-recipes benchmarks to validate AI agent improvements against baseline models, ensuring performance and accuracy.

Architects high-precision AI agent systems using Stagehand V3, enabling proactive code research and resilient automation.

Comprehensive tool for spreadsheet creation, editing, analysis, and visualization with formula support and intelligent data import capabilities.

Translates PyTorch implementations to Triton GPU kernels incrementally for optimization, auditing, and AI model acceleration.

Provides best practices for crafting effective prompts to optimize interactions with AI agents and Claude, enhancing response quality and reliability.

Enables multi-agent coordination via Synapse A2A framework, handling routing, priority, task history, and file safety for agent communication.

Provides an iterative agent pattern with orchestrator, worker, and judge for refining outputs until acceptance criteria are met within a defined budget.

Generates and executes database migration scripts to safely alter schemas while preserving data integrity.

Assists in working with Lean 4 quotients (Quot.lift, Quot.ind, Quot.sound) and nested lifts for formal proof obligations.

Automates filling 'sorry' placeholders in Lean files using the Aristotle theorem prover, handling API setup, HIT detection, and result verification.

Switches OpenEvent-AI between OpenAI, Gemini, hybrid, or stub LLM providers via environment variables or admin API endpoints.

Generates AI voiceovers for Remotion videos using Resemble.ai's Text-to-Speech technology.

Provides a PostgreSQL-based memory system for AI agents to store, retrieve, and search contextual data from past sessions and external sources.

Debugs misclassifications in OpenEvent-AI's intent detection and routing, resolving LLM vs regex conflicts and trigger issues.

Provides agent skills for context engineering, multi-agent system design, and debugging in production AI agent environments.

Transcribes YouTube video audio to text or JSON via Deepgram, with optional AI summary generation using Claude.

Automates path_simp, path_auto, and path_normalize tactics for structured calc-based proofs in ComputationalPaths.

Converts scanned documents and images to Markdown format using Mistral AI's OCR capabilities. Ideal for processing PDFs and image-based documents.

Extracts clean Markdown content and metadata (title, author, date) from web pages with dry-run verification and JSON output.

Manages the system prompt and greeting for the Vocal Bridge voice agent, enabling viewing, editing, and configuration.

Optimizes AI agent memory and token usage by compressing conversation history in long sessions, reducing context length without significant information loss.