AI in the Trenches: Engineering Real-World Systems
summary
Gartner prediction that 30% of GenAI projects fail by end of 2025 anchors a practical agenda: build data platforms, measure business KPIs, and use retriever-augmented generation plus guardrails and observability to cut hallucinations. Concrete wins include a RAG chatbot indexing thousands of PDFs for researchers and a product team that ships 22 integrations in 3 weeks versus months previously. Expect concrete engineering advice on testing, continuous evals, context sizing, agent design (planning/reasoning/tools/autonomy), and tradeoffs like energy costs vs. model size.
AI-written summary · about
chapters (10)
- 0:00Introduction and panel setup
- 1:04AI project failure predictions
- 10:58Business value and AI strategy
- 20:21Consumer AI successes and failures
- 28:52When to use generative AI
- 36:03Handling AI hallucinations
- 43:52Defining AI agents
- 48:05AI's environmental impact
- 54:28AI in autonomous systems
- 1:04:44Closing and key takeaways
related talks
28mIs GenAI the technology revolution we need? How can we harness it at work?
50mAI IN SRE: HYPE VS REALITY IN 2026
31mGenAI after the Hype: Transforming Organizations with GenAI-based Agents
27mSoftware engineering with GenAI
16mBuilding Applications with AI Agents
17mGenerative AI: Keeping it real
45mThree Practical Ways to Use AI in Your Company
23mAI and APIs in 2026