Complete procedures with a way back.

Build an educational causal language model from randomly initialized weights using a licensed, provenance-tracked corpus. Train a dedicated tokenizer, define a roughly 34-million-parameter decoder architecture, establish random and unigram baselines, checkpoint and resume one bounded epoch, evaluate held-out loss and qualitative limitations, run local inference, and publish only an honest research artifact.

Build a governed supervised fine-tuning pipeline for a small open-weight causal model: prove base and dataset licenses, audit immutable splits, establish a frozen baseline, train a LoRA or QLoRA adapter, resume checkpoints, evaluate regressions, merge deliberately, test post-training quantization, and serve a pinned adapter through an authenticated API.

Instrument complete agent trajectories with privacy-safe OpenAI Agents SDK traces and OpenTelemetry, correlate model calls, handoffs, guardrails, approvals, tools, retries, and side effects, measure end-to-end latency and all token cost, evaluate unsafe behavior deterministically, and release dashboards and alerts with the agent.

Design agent working memory, persistent conversation state, compaction, and durable facts as explicit governed application systems with tenant isolation, confirmation, provenance, minimization, TTLs, correction, verified deletion, evaluation, cost monitoring, and rollback that cannot resurrect prohibited memories.

Build a production RAG evaluation program that separates retrieval recall and precision from answer faithfulness, citation validity and support, correctness, usefulness, and abstention; calibrates automated graders with humans; blocks security regressions; and carries evidence through regression, canary, continuous evaluation, and rollback.

Design and release a production RAG retrieval pipeline that preserves exact-term BM25 strength, adds semantic vector recall, fuses rankings with reciprocal rank fusion, reranks a bounded set, applies tenant ACLs before every branch, and proves quality, latency, cost, citation, and rollback behavior.

Govern source revisions, preserve ACLs through parsing and chunking, pin embeddings, combine lexical and semantic retrieval, validate claim-level citations, evaluate abstention, and release immutable indexes with safe rollback.

Build an evidence-first support bot that crawls one authorized website through sitemap and robots controls, blocks SSRF, normalizes canonical pages, creates an OpenAI File Search index, returns inspectable citations, evaluates every revision, exposes an explicit human-escalation path, and refreshes content without replacing a known-good index.