Applied AI, Artificial Intelligence & Machine Learning

Building an LLMOps Pipeline: Prompt Versioning, Drift Monitoring, and CI/CD

May 26, 2026

Prompt versioning CLI, LLM-as-judge eval harness, PSI drift monitoring, GitHub Actions CI/CD with a hard quality gate, and a Streamlit monitoring dashboard.

LoyaltyLens Series Recap: Architecture Decisions, Lessons, and What to Build Next

May 21, 2026

Module-by-module architecture decisions, five key lessons from building a production AI system end-to-end, and a prioritised roadmap for extending LoyaltyLens to production.

Wiring Six AI Modules Into One Pipeline — and Deploying to the Cloud

May 19, 2026

How I connected LoyaltyLens's six modules into a single orchestration layer, exported the PyTorch model to TorchScript, and deployed a real-time inference endpoint on AWS SageMaker.

Building a Feedback Loop for Production LLMs: Signals, Preference Data, and Retraining Triggers

May 14, 2026

Closing the production AI loop — feedback capture API, React review UI, preference dataset export, automated retraining triggers, and the practical limits of RLHF without a research team.

#rlhf

#reinforcement-learning

RAG Offer Retrieval: LangChain vs. LlamaIndex, pgvector vs. Weaviate

May 12, 2026

Building a RAG offer retrieval system — latency and precision benchmarks comparing LangChain vs. LlamaIndex and pgvector vs. Weaviate at three catalog sizes.

LLM Offer Copy Generation: Prompt Registry, Dual Backends, and Brand Image Generation

May 7, 2026

Building the LLM offer copy generator — versioned YAML prompt registry, dual LLM backends, JSON parse retry logic, and FLUX.1 brand image generation via HuggingFace.