Machine learning

Use when building AI workflows, tool calling agents, structured outputs, or LLM pipelines using the Genkit Dart SDK.

Generate system prompts from user requirements, with application guidance and request-derived evaluation materials.

Remove refusal behaviors from open-weight LLMs using OBLITERATUS — mechanistic interpretability techniques (diff-in-means, SVD, whitened SVD, LEACE, SAE…

Configure Azure API Management as an AI Gateway for AI models, MCP tools, and agents.

Use when a task gives a covariate table plus, for each of many simulated observational datasets, a binary treatment vector and an outcome vector, and asks for…

Push the LLM to reconsider, refine, and improve its recent output.

Benchmark stellar-build skills to measure quality and the lift from optimization.

Tend the membrane memory corpus — an OKF concept store whose fragment evolution lives in the git log as lineage trailers.

Design and analyze engineering experiments or simulations using baseline cases, hypothesis-driven DOE, traceable data processing, defensible plots…

Answer Engine Optimization (AEO): optimize content to be cited by LLMs (ChatGPT, Claude, Perplexity, Gemini) in their answers.

Context management engine for AI coding agents.

Data science across machine learning, statistical modeling, and experimentation.

MLOps across model deployment, ML pipelines, monitoring, and feature stores.

Design RAG pipelines: chunking, retrieval evaluation, and architecture.

Computer vision engineering for object detection, segmentation, and visual AI, covering CNN and Vision Transformer architectures and ONNX/TensorRT deployment.

Prompt engineering and LLM evaluation.

Stop an AI agent damaging your repo: PreToolUse hooks, permission deny rules, protected paths, verification gates.

Two-phase security audit of the indexed group.

AI features you ship to users: structured output, tool schemas, prompt injection, evals.

Best practices for dataset preparation and PEFT/LoRA fine-tuning.

Turn a bug, outage or repeated mistake into a device that makes the whole class impossible.

Expert guidance for Fully Sharded Data Parallel training with PyTorch FSDP - parameter sharding, mixed precision, CPU offloading, FSDP2

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart".

Use for "why does X work this way", "why did we choose Y", design rationale, regressions, postmortems, historical constraints, or data-backed thresholds.