3. Data, Science & AI

Automates Google NotebookLM for AI-generated content creation including podcasts, videos, and quizzes from user-provided sources.

Updates AgentDB vector store with new case patterns while monitoring for bias in AI model training data.

Quick reference guide for Plotly Dash BI dashboard commands including startup, shutdown, verification, and troubleshooting procedures.

Guides implementation of Bronze/Gold layered data architecture for LLM-extracted data, managing extraction history and human-verified corrections.

Provides optimized query patterns and data engineering best practices for ClickHouse to enhance analytical workload performance.

Isolates skin regions from images using threshold-based segmentation for downstream analysis in medical or scientific applications.

Retrieves 10 similar historical cases from AgentDB vector store to enable RAG-based diagnosis support in medical or analytical contexts.

Generates actionable clinical advice calibrated for patient skin tone using AI models (WebLLM or Gemini).

Classifies patient skin tone using Fitzpatrick/Monk scale with Gemini 3 Flash AI, outputting confidence score for medical assessment.

Enables real-time medical literature search to validate clinical diagnoses using Google Search Tool.

Analyzes video visibility using head pose, face mesh, and temporal patterns to differentiate face-visible from body-only segments.

AI splits complex tasks, concurrently executes code via Codex, and aggregates results for review.

Supports BAML (Boundary ML) integration in Sagebase, enabling definition file creation, client regeneration, and Factory Pattern implementation for hybrid AI systems.

Validates demographic fairness in AI models by analyzing True Positive Rate and False Positive Rate gaps using AgentDB metrics.

Performs face alignment, landmark extraction, and quality assessment to debug alignment issues and filter facial images.

Generates comprehensive medical risk profiles by synthesizing lesion data and historical cases using the SmolLM2 language model.

Performs advanced econometric analysis including regression, causal inference, and time series for empirical economic research.

AI-powered diagnostics for computer vision pipelines, explaining face detection, tracking, embedding, and clustering failures using GPT-4o analysis.

Optimizes embedding model backends (InsightFace, PyTorch, ONNXRuntime, TensorRT) for improved throughput and debugging of drift/fallbacks.

Explains Polibase data processing workflows including meeting minutes, web scraping, and politician data collection with dependency and execution order details.

Enables person tracking and re-identification using body features without facial data, for consistent multi-camera tracking.

Runs MOSAIC microsimulation and extracts formatted results for academic papers.

Extracts structured data from webpages, enabling users to scrape and retrieve specific information from URLs for data analysis.

Extracts images, videos, and documents embedded in BSV blockchain transactions using the txex CLI tool to retrieve inscribed ordinals.