Original frameworks, architectures, and field notes from 28 years of shipping AI to production. Every artifact below is open-source, code-first, and battle-tested before it was named.
Production AI evaluation and security primitives we've designed, implemented, and released as open source.
Framework · 2025A framework for scoring LLM responses across Groundedness, Accuracy, Reliability, Variance, Inference cost, and Safety. The open-source implementation (Beta, Apache-2.0) uses rule-based heuristics, not trained models, and has no benchmarks against labelled data. Weight profiles for enterprise / healthcare / finance / consumer / agentic only change how scores are weighted. Install with pip install argus-llm.
Three formula-based metrics for monitoring autonomous workflows from step counts, retries, and cost: Agent Stability Factor, Error Recovery Rate, Cost Per Completed Step.
An end-to-end agent security architecture: input sanitizer (zero-permission isolate) → injection detector (BERT + pattern catalog) → compartmentalized RBAC (zero trust for agents) → human confirmation gate (async, multi-channel) → encrypted audit trail (Fernet, 7-year retention). MCP-native, vendor-neutral. Released as pip install bulwark-agent-security.
Long-form writing on production AI engineering — what the failure modes actually look like, and how the architectural choices land.
claude-auth-setup open-source release.Subscribe on Medium for new field notes on production agentic AI, evaluation, and security.
Subscribe on Medium →Every framework, every architecture, every paper — code-first, Apache 2.0, on GitHub.
OSS · Apache 2.0Five-layer agent security framework. Sanitizer + detector + RBAC + audit + human gate. pip install bulwark-agent-security
Reference implementation of G-ARVIS (Beta, Apache-2.0). Heuristic six-dimension LLM scoring plus agentic metrics. pip install argus-llm
MLOps platform for the production model lifecycle: experiment tracking, model registry, automated evaluation pipelines, drift detection, governance dashboards.
PyTorch production framework — standardized training loops, model registry, one-command deployment. Built because most PyTorch models never reach production.
Reproducible, evaluation-first orchestration for LLM fine-tuning — RLAIF, LLM-as-Judge, a data flywheel, RAG generation, adapter merging, and benchmark evaluation behind a REST API.
Explainable AI framework for regulated industries — human-readable explanations with audit-grade traceability for healthcare, finance, and compliance environments.
Vitess-based horizontal MySQL sharding for healthcare and genomics. Zero-downtime migrations from monolithic to sharded production.
Cross-platform Bash + Batch scripts for Claude Code authentication setup. Six auth paths (subscription, API key, OAuth, AWS Bedrock, Google Vertex AI, Microsoft Foundry) behind one guided flow. Production-tested on Windows, macOS, Linux, WSL.