Workshops: Agentic Coding by Tech Stack
LLM App Engineering with Coding Agents: 2-Day Training
An LLM feature is easy to demo and hard to run reliably: outputs vary, costs grow with usage and new attack surfaces appear. In this two-day training, product teams build a real LLM feature with coding agents as their pair and learn the engineering practices that make it production-ready: structured outputs, tool calling, retrieval, evals, guardrails and cost control.
Who is this workshop for?
- Software engineers adding LLM features to existing products
- AI and ML engineers moving from prototype to production
- Tech leads and architects choosing models, providers and hosting
- Product teams who own the quality of AI features
- Platform teams building shared LLM gateways or SDK wrappers
What you will take away
- Integrate model APIs via official SDKs and validate structured outputs against schemas.
- Design tools for function calling and implement agentic loops with clear limits.
- Build a retrieval pipeline and recognise when retrieval actually improves answers.
- Set up an eval suite in CI that catches regressions when prompts or models change.
- Apply guardrails, observability and cost controls, and know the relevant transparency duties under the EU AI Act.
Agenda
Module 1
Day 1: Foundations and architecture
- APIs and SDKs from Anthropic, OpenAI and Mistral, plus cloud offerings such as Amazon Bedrock, Azure and Google Vertex AI
- Claude Code and other coding agents as a pair: what to delegate in LLM projects
- Exercise kick-off: the feature you will build over two days
Module 2
Day 1: Structured outputs and tools
- JSON schemas, validation and retries
- Tool calling: designing tools, handling errors, limiting loops
- Streaming and UX for slow or uncertain answers
Module 3
Day 1: Retrieval
- Chunking, embeddings and hybrid search, for example with pgvector
- When long context beats retrieval
- Citations and source handling
Module 4
Day 2: Evals
- Building a golden dataset from real use cases
- Code-based checks and calibrated LLM-as-judge
- Eval runs in CI as a regression gate
Module 5
Day 2: Guardrails and security
- Prompt injection, data exfiltration via tools, output filtering
- Personal data, logging and retention
- Human in the loop for consequential actions
Module 6
Day 2: Operations and cost
- Tracing and observability, for example with OpenTelemetry
- Prompt caching, model routing and token budgets
- A release process for prompt and model changes
Agents build, evals decide
Coding agents excel at the plumbing of LLM features: SDK integration, schemas, retry logic, test harnesses and dashboards. They are less reliable at judging whether the feature actually serves users well. That is why evals sit at the centre: the agent builds quickly, and a measurable eval suite decides whether a change is an improvement.
You leave with a working feature skeleton, an eval set and a release checklist you can reuse for the next feature.
Regulation and data protection
LLM features often process personal data, and some fall under transparency duties of the EU AI Act, for example when people interact directly with an AI system or when AI-generated content must be marked. We explain the requirements, data processing agreements, EU hosting options and data minimisation in prompts and logs. The legal assessment stays with your legal and compliance team.
Model-agnostic by design
We work with Claude, OpenAI models, Mistral and open-weight models, depending on your requirements. Provider-specific code stays behind a thin abstraction, so you can switch when quality, cost or data residency demand it. Sessions are led by neooutput AI Code Engineers and AI Solutions Architects, remotely or on-site, in English or German.
Frequently asked questions
Do participants need machine learning experience?
No. Solid software engineering skills are enough. We work at the level of APIs, data and tests, not model training.
Which frameworks do you use?
We start close to the official SDKs because that keeps behaviour transparent. LangChain, LlamaIndex or the Vercel AI SDK come in where they fit your stack, and we discuss the trade-offs.
Can we work on our own use case?
Yes, and we recommend it. Ideally you bring a feature idea with sample data that may be used in the training; otherwise we provide a realistic scenario.
Workshop
Get in touch
Want your next AI feature to be measurable, safe and affordable from day one?
Tell us roughly how many people will join and pick dates that work – we will get back to you within 24 hours with a proposal for content, format and schedule.
Prefer to talk first?
15 minutes, free and without obligation – with our AI Trainers and AI Code Engineers.
Book a call for free (opens in new tab) +49 761 519 751 50 (24/7)Request a training / workshop
Workshop: LLM App Engineering with Coding Agents: 2-Day Training · 2 days
More workshops
- Workshops: Agentic Coding by Tech StackEmbedded C/C++ & Firmware with AI Agents: Workshop
- Workshops: Agentic Coding by Tech StackCOBOL & Mainframe Modernisation with AI Agents: Workshop
- Workshops: Agentic Coding by Tech StackWordPress Plugin & Theme Development with AI Agents: Workshop
- Workshops: Agentic Coding by Tech StackShopify Themes & Apps with AI Coding Agents: Workshop
- Workshops: Agentic Coding by Tech StackGame Development with AI Coding Agents: Unity, Unreal, Godot
- Workshops: AI by Department & IndustryAI in Procurement & Supply Chain Workshop
