3. Data, Science & AI

Provides a framework for classifying data by sensitivity levels, including labeling, handling requirements, and lifecycle management.

Configures PITER framework for autonomous agent systems, including prompt inputs, triggers, environments, and review processes in agent workflows.

Optimizes AI prompt engineering by reducing context size, delegating tasks to specialized agents, and managing context systematically for efficient LLM usage.

Tracks and measures key performance indicators for agentic coding workflows, including size, attempts, streak, and presence metrics.

Optimizes machine learning inference performance via model compression, distillation, and edge deployment strategies to reduce latency and size.

Guides selection and implementation of ETL/ELT data pipeline patterns for modern data architecture and transformation needs.

Analyzes gene annotations, TSS, and transcript data using GENCODE and GTF formats for genomic region queries.

Analyzes expression data to infer causal relationships and visualize gene regulatory networks using LiNGAM algorithm.

Creates domain-specific AI agents with pre-loaded knowledge using the Act-Learn-Reuse pattern to maintain mental models and enable self-improvement via expertise files.

Executes gene set and pathway enrichment analysis (GO, KEGG, Reactome) via gcell for genomic data interpretation.

Evaluates agentic layer maturity in codebases using a 12-grade classification system (Classes 1-3) for readiness assessment and upgrade planning.

Analyzes protein structures, sequences, and interactions using AlphaFold2 predictions, UniProt data, and STRING database networks with 3D visualization.

Guides design, refactoring, and optimization of single-purpose AI agents using the One Agent One Prompt One Purpose principle.

Designs modular, reusable sections for constructing agentic prompts, enabling structured and composable prompt engineering for AI agent workflows.

Creates domain-specific AI agents with full SDK control, custom tools, and tailored prompts using Claude Agent SDK patterns.

Analyzes cell type regulatory relationships using gcell, featuring gene-by-motif matrices, Jacobian analysis, and motif subnet visualization for gene expression patterns.

Analyzes DNA sequences, scans motifs with HOCOMOCO, and processes genomic regions (BED, narrowPeak) using genome assemblies (hg38, hg19, mm10).

Designs orchestrator systems for managing AI agent fleets, including multi-agent architectures and unified control interfaces.

Generates prompts to align agent expertise with current codebase, enabling self-improvement in maintenance workflows and the LEARN phase of Act-Learn-Reuse.

Designs system prompts to define agent identity, rules, and behavior for custom AI agents, ensuring effective task execution and interaction.

Manages agent fleet lifecycle via CRUD operations, monitoring, and resource cleanup in multi-agent AI systems.

Analyzes Excel spreadsheets to generate visual charts for data representation. Use when processing .xlsx files or performing spreadsheet analysis.

Audits AI model context composition to identify token usage inefficiencies and optimize context window utilization, enhancing agent performance.

Parses Gemini CLI output in JSON and stream-JSON formats, handling response extraction, stats, errors, and tool calls programmatically.