> Markdown version of [/events/world-congress-2026-europe/sessions/1094-tour-de-force-open](https://www.wearedevelopers.com/events/world-congress-2026-europe/sessions/1094-tour-de-force-open). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Tour de Force: Open-Source LLM Inference Optimization from Simple to Sophisticated


- **Date:** Thursday, Jul 9, 2026
- **Time:** 14:10–14:40 (30 min)
- **Room:** Stage 6 - powered by Microsoft
- **Event:** World Congress 2026 Europe

## Recording

[Watch recording](https://www.wearedevelopers.com/videos/100099-tour-de-force-open-source-llm-inference-optimization-from-simple-to-sophisticated)

## Description

Azure OpenAI and similar managed APIs are the right default for serving language models. But they don't cover every case. Maybe you need to deploy in a region where your model isn't available yet, you want to run a Qwen or Mistral variant that no provider hosts, or you've fine-tuned a model and there's simply no API to call. At that point, you're self-hosting on GPUs.

Making your GPUs go brrr is complex. Efficient LLM inference requires navigating a maze of optimization techniques each with different trade-offs. This session provides a practical journey through inference optimizations, clearly categorized by implementation effort.

We'll explore techniques across three levels:

- Model choices (start here): Model selection, quantization, smart routing

- Library-level improvements (using PyTorch-based frameworks like vLLM, SGLang, TensorRT-LLM): Continuous batching, KV-cache management

- Custom implementations: Speculative decoding with custom draft heads, disaggregated inference, fine-tuning smaller models

The session covers practical trade-offs and key metrics: time to first token, inter-token latency, and cost per token.

Whether deploying your first model or optimizing at scale, this talk delivers actionable insights into which techniques to prioritize for deeper investigation.

## Speaker

### [Christin Pohl](https://www.wearedevelopers.com/@christin-pohl)

Global Black Belt Solution Engineer at Microsoft


## Related talks at this congress

- [Owning the Inference Layer: When and How to Run your Own Models](https://www.wearedevelopers.com/events/world-congress-2026-europe/sessions/983-owning-the-inference) — Taylor Jordan Smith
- [Faster Together: Train and Deploy a Speculative Decoding Model for Low-Latency LLM Inference](https://www.wearedevelopers.com/events/world-congress-2026-europe/sessions/1142-faster-together) — Amit Kushwaha, Dmitry Mironov
- [Agents That Own Their Inference: Building Production AI Agents on Dedicated GPUs](https://www.wearedevelopers.com/events/world-congress-2026-europe/sessions/1350-agents-that-own) — Duan Lightfoot
- [The LLM Evolution: From Sequence Imitation to Verifiable Reasoning](https://www.wearedevelopers.com/events/world-congress-2026-europe/sessions/1377-the-llm-evolution) — Kamen Petroff