Conf42 Prompt Engineering 2026
who it's for
prompt engineers, AI safety engineers, compliance engineers, agent developers, machine learning engineers, observability engineers
topics
evals as unit tests; prompt injection testing; context engineering for long tokens; personalized RAG for enterprise analytics; multi-layer guardrail architectures; prompt drift detection and stability; Terraform change simulation for infrastructure; human-in-the-loop compliance workflows; agent observability with Datadog; bias detection with owned models
| title | author | length |
|---|---|---|
| Your Agent Passed the Prompt Test. Will It Survive Production? | Nicolas Zelaya | 22m |
Harness shows a reproducible pipeline: run evals as unit tests against a golden dataset and 50 built-in metrics, deploy variations via AI configs/FME as a… | ||
| Prompting Is Not the System: Context, Contracts, Reliable Agent | Emerson Braun | 18m |
A content pipeline is built to auto-create posts but an empty draft slips into production because the prompt alone is insufficient; the concrete fix splits the… | ||
| Whose gate is it? | Ian Johnson | 15m |
Claim: split the agent harness into charter, harness, factory with one-way dependencies (factory → harness → charter) so rules live where they belong. Defines… | ||
| Your Agent Can Write Terraform. It Can't Feel What It Breaks. | Kevin Brown | 09m |
A GitHub app called Cloud World Model GitHub app runs a simulation on Terraform pull requests and inserts a report showing modeled cost, P99 latency, CPU… | ||
| Prompt Engineering for AI-Powered NEPA Compliance Automation | Mohit Upadhyay | 29m |
Claims that permitting is primarily an evidence-first design problem not a document one, and that prompts must act as controlled instructions that forbid… | ||
| Prompt Engineering for Agentic Safety Assurance in Physical AI | Vraj Patel | 14m |
Argues that prompts alone fail for embodied robots and proposes four pillars anchored by a system graph that encodes ODDs, mission scenarios, requirements and… | ||
| Prompt Engineering for AML: Build Reliable Compliance Workflows | Pratik Patel | 26m |
Explains how prompts become bank controls by using a five-layer architecture (data, cloud platform, prompt orchestration, compliance apps, governance) and… | ||
| Single-Layer LLM Guardrails Fail: 390 Prompt Injection Tests | Manvitha Potluri | 23m |
A 390‑prompt evaluation (172 single‑turn attacks, 18 multi‑turn attacks, 200 legitimate prompts) shows a single-layer guardrail with keyword rules catches… | ||
| Context Engineering for MCP: 150K-22K Tokens | Nishant Kumar Thakur, Sonam Raj | 20m |
Measured context goes from 1.17M tokens for 2,594 tools down to a single-code approach that often needs ~1k tokens; a production agent moves from 150K to 22K… | ||
| What a Front-Facing Agent Taught Us About Prompting | Ahmed Gaber, Ahmed Morsi | 35m |
A front-line hybrid copilot called Cobalt handles customer chats, searches prices, drafts replies and hands tasks to reps; after cleaning data with one ID for… | ||
| From Prompt to Product | LaShawnda Rodgers | 51m |
Validated idea-to-launch recipe: problem discovery with Google Trends and Reddit, research citations from Perplexity, then a potty-training app and a senior… | ||
| Detecting 70 Kinds of Bias with a Model You Own: Replacing LLMs | Aleem Khan | 25m |
Collects news from X and Google News, deduplicates and threads stories, then queries OpenAI, Anthropic, Quen, Croak and Deep Seek with an expert-written… | ||
| Observing AWS AgentCore with Datadog: End-to-End O11y | Indika Wimalasuriya | 33m |
An end-to-end demo wires AWS AgentCore on Bedrock (Claude Sonnet 4.4/4.5) through the Strand agent SDK into Datadog APM and LLM Observability, sending OTLP to… | ||
| Beyond Better Prompts: Prompt Engineering for Production AI | Emmanuela Opurum | 31m |
Argues that every AI output requires evaluation against concrete project standards, shows using Claude, Codex with project rule files (.md) and saved prompt… | ||
| You Can't Fix a Prompt You Can't See Fail | Krishna Chaitanya Balusu | 14m |
Shows a concrete fix: wrap the planning step in an OpenTelemetry plan span that stores the list of tools, then run a separate answer check and a plan check… | ||
| Prompt Drift Kills: How to Make Prompts Stable | Sachin Gupta | 14m |
Demo builds an IT assistant that must state debug logs retained for 14 days and return JSON fields answer,status,citation,exact_source; checks run as Java CI… | ||
| Personalizing RAG for Prompt-Driven Enterprise Analytics | Santosh Muthyala | 23m |
RAG-based system personalizes presentation while keeping facts constant: a concrete example shows checkout conversion falling 5% to 4.5% with mobile driving… | ||
| Beyond Basic RAG: Better Context for More Reliable Enterprise AI | Ravi Ray | 14m |
A vendor-neutral five‑axis framework (ingestion, indexing, retrieval composition, grounding, governance) guides moving RAG from prototype to production, based… | ||
| nothing matches | ||