{"schemaVersion":1,"id":"https://lucioduran.com/speaking","url":"https://lucioduran.com/speaking","title":"Artificial intelligence talks and training","description":"Talk and workshop proposals by Lucio Durán in Uruguay: AI agents, LLM evaluation, RAG and inference, grounded in open-source projects and reproducible evidence.","language":"en","datePublished":"2026-11-10","dateModified":"2026-11-10","contentFormat":"text/markdown","content":"# Artificial intelligence talks and training\n\nI share technical work on artificial intelligence from Uruguay. These proposals are for communities, universities, organizations and teams that want to understand how to design and evaluate AI systems using code and evidence.\n\nThe starting point is my public projects: architecture decisions, experiments, negative results and limitations. Each topic can become a talk about decision criteria or a workshop that examines an implementation and its results.\n\n[Invite me to speak or propose a training session](<https://lucioduran.com/speaking#invitation>)\n\n## Proposed topics\n\n### AI agents that act: permissions, failures and recovery\n\n**Audience:** Software and AI teams, developer communities and students with programming fundamentals.\n\n**What we explore**\n\n- Separate the model’s proposal from tool authorization and execution.\n- Analyze a timeout without assuming the operation failed or repeating its effect.\n- Examine traces, permissions and recovery tests in the reference projects.\n\n**Public reference material:**\n\n- [Agent Replay Runtime](<https://lucioduran.com/projects/agent-replay-runtime>) · [Source code](<https://github.com/lucioduran/agent-replay-runtime>)\n- [Agent Tool Capabilities](<https://lucioduran.com/projects/agent-tool-capabilities>) · [Source code](<https://github.com/lucioduran/agent-tool-capabilities>)\n\n[AI agent architecture](<https://lucioduran.com/topics/ai-agents>)\n\n### Evaluate models and agents before accepting an improvement\n\n**Audience:** AI and data teams, quality specialists, and technology or product leaders.\n\n**What we explore**\n\n- Define a task, acceptance criteria and a versioned evaluation set.\n- Distinguish task success, constraint compliance and regressions by segment.\n- Interpret results, failures and cases requiring review instead of automatic acceptance.\n\n**Public reference material:**\n\n- [Model Regression Gate](<https://lucioduran.com/projects/model-regression-gate>) · [Source code](<https://github.com/lucioduran/model-regression-gate>)\n- [Agent Injection Evals](<https://lucioduran.com/projects/agent-injection-evals>) · [Source code](<https://github.com/lucioduran/agent-injection-evals>)\n\n[LLM and AI agent evaluation](<https://lucioduran.com/topics/llm-evaluation>)\n\n### RAG with evidence: retrieval, languages and answers\n\n**Audience:** Teams building document search or assistants, and engineering or data science students.\n\n**What we explore**\n\n- Separate retrieval quality from answer quality.\n- Compare lexical, dense and hybrid approaches on a defined corpus and query set.\n- Examine ranking errors and differences between English and Spanish queries and documents.\n\n**Public reference material:**\n\n- [RAG Evaluation Workbench](<https://lucioduran.com/projects/rag-evaluation-workbench>) · [Source code](<https://github.com/lucioduran/rag-evaluation-workbench>)\n- [Bilingual Retrieval Evals](<https://lucioduran.com/projects/bilingual-retrieval-evals>) · [Source code](<https://github.com/lucioduran/bilingual-retrieval-evals>)\n\n[RAG: retrieval, search and evaluation](<https://lucioduran.com/topics/rag>)\n\n### AI inference: measure latency, memory and quality\n\n**Audience:** Platform engineers, AI teams and people making infrastructure decisions.\n\n**What we explore**\n\n- Distinguish first-token wait, generation time and total duration.\n- Measure weight storage and process memory as separate variables.\n- Read a quantization experiment without extrapolating its results to another model or hardware.\n\n**Public reference material:**\n\n- [LLM Serving Bench](<https://lucioduran.com/projects/llm-serving-bench>) · [Source code](<https://github.com/lucioduran/llm-serving-bench>)\n- [Quantization Quality Lab](<https://lucioduran.com/projects/quantization-quality-lab>) · [Source code](<https://github.com/lucioduran/quantization-quality-lab>)\n\n[AI inference: latency, memory and quality](<https://lucioduran.com/topics/llm-inference>)\n\n## Formats to discuss\n\nTechnical talk\n\nA presentation centered on a problem, the decisions that shape it and the available evidence, with time for questions.\n\nCode workshop\n\nA walkthrough of a public repository, its configuration and its tests. Scope depends on the group’s experience, time and available environment.\n\nSession for technology leaders\n\nA discussion of criteria for evaluating an AI proposal: what to test, which risks to control and which evidence to request before deciding.\n\n## How a session is prepared\n\n1. We agree on the audience, prior experience and the decision or skill the session should help develop.\n2. We select a topic and public materials that fit the scope. For a workshop, we define technical requirements and what will be run or reviewed.\n3. We distinguish observations from the experiments from claims that still need validation in another environment. References let participants continue working after the session.\n\n## Short introduction for organizers\n\nLucio Durán is C.T.O. at LOi and works from Uruguay. His public work combines software architecture with applied artificial intelligence, including open-source projects on agents, language model evaluation, RAG and inference. He publishes code, methods and results alongside each experiment’s scope and limitations.\n\n[Experience and education](<https://lucioduran.com/cv>) · [All projects](<https://lucioduran.com/projects>)\n\n## Propose an invitation\n\nSend the organization or event name, audience, topic of interest, tentative dates, in-person or remote format and available time. With that context, we can discuss scope, requirements and terms.\n\n[Send a proposal](<mailto:email@lucioduran.com?subject=Invitation%3A%20artificial%20intelligence%20talk%20or%20training&body=Organization%20or%20event%3A%0AAudience%20and%20prior%20experience%3A%0ATopic%20of%20interest%3A%0ATentative%20dates%3A%0AFormat%20and%20location%3A%0AAvailable%20time%3A%0ASession%20objective%3A%0A>) · [email@lucioduran.com](<mailto:email@lucioduran.com>)\n\n[Other ways to get in touch](<https://lucioduran.com/contact>)","headings":[{"level":1,"text":"Artificial intelligence talks and training","url":"https://lucioduran.com/speaking#title"},{"level":2,"text":"Proposed topics","url":"https://lucioduran.com/speaking#proposed-topics"},{"level":3,"text":"AI agents that act: permissions, failures and recovery","url":"https://lucioduran.com/speaking#ai-agents"},{"level":3,"text":"Evaluate models and agents before accepting an improvement","url":"https://lucioduran.com/speaking#llm-evaluation"},{"level":3,"text":"RAG with evidence: retrieval, languages and answers","url":"https://lucioduran.com/speaking#rag"},{"level":3,"text":"AI inference: measure latency, memory and quality","url":"https://lucioduran.com/speaking#llm-inference"},{"level":2,"text":"Formats to discuss","url":"https://lucioduran.com/speaking#formats"},{"level":2,"text":"How a session is prepared","url":"https://lucioduran.com/speaking#preparation"},{"level":2,"text":"Short introduction for organizers","url":"https://lucioduran.com/speaking#speaker-bio"},{"level":2,"text":"Propose an invitation","url":"https://lucioduran.com/speaking#invitation"}],"sections":[{"heading":"Artificial intelligence talks and training","level":1,"url":"https://lucioduran.com/speaking","content":""},{"heading":"Artificial intelligence talks and training","level":1,"url":"https://lucioduran.com/speaking#title","content":"I share technical work on artificial intelligence from Uruguay. These proposals are for communities, universities, organizations and teams that want to understand how to design and evaluate AI systems using code and evidence.\n\nThe starting point is my public projects: architecture decisions, experiments, negative results and limitations. Each topic can become a talk about decision criteria or a workshop that examines an implementation and its results.\n\n[Invite me to speak or propose a training session](<https://lucioduran.com/speaking#invitation>)"},{"heading":"Proposed topics","level":2,"url":"https://lucioduran.com/speaking#proposed-topics","content":""},{"heading":"AI agents that act: permissions, failures and recovery","level":3,"url":"https://lucioduran.com/speaking#ai-agents","content":"**Audience:** Software and AI teams, developer communities and students with programming fundamentals.\n\n**What we explore**\n\n- Separate the model’s proposal from tool authorization and execution.\n- Analyze a timeout without assuming the operation failed or repeating its effect.\n- Examine traces, permissions and recovery tests in the reference projects.\n\n**Public reference material:**\n\n- [Agent Replay Runtime](<https://lucioduran.com/projects/agent-replay-runtime>) · [Source code](<https://github.com/lucioduran/agent-replay-runtime>)\n- [Agent Tool Capabilities](<https://lucioduran.com/projects/agent-tool-capabilities>) · [Source code](<https://github.com/lucioduran/agent-tool-capabilities>)\n\n[AI agent architecture](<https://lucioduran.com/topics/ai-agents>)"},{"heading":"Evaluate models and agents before accepting an improvement","level":3,"url":"https://lucioduran.com/speaking#llm-evaluation","content":"**Audience:** AI and data teams, quality specialists, and technology or product leaders.\n\n**What we explore**\n\n- Define a task, acceptance criteria and a versioned evaluation set.\n- Distinguish task success, constraint compliance and regressions by segment.\n- Interpret results, failures and cases requiring review instead of automatic acceptance.\n\n**Public reference material:**\n\n- [Model Regression Gate](<https://lucioduran.com/projects/model-regression-gate>) · [Source code](<https://github.com/lucioduran/model-regression-gate>)\n- [Agent Injection Evals](<https://lucioduran.com/projects/agent-injection-evals>) · [Source code](<https://github.com/lucioduran/agent-injection-evals>)\n\n[LLM and AI agent evaluation](<https://lucioduran.com/topics/llm-evaluation>)"},{"heading":"RAG with evidence: retrieval, languages and answers","level":3,"url":"https://lucioduran.com/speaking#rag","content":"**Audience:** Teams building document search or assistants, and engineering or data science students.\n\n**What we explore**\n\n- Separate retrieval quality from answer quality.\n- Compare lexical, dense and hybrid approaches on a defined corpus and query set.\n- Examine ranking errors and differences between English and Spanish queries and documents.\n\n**Public reference material:**\n\n- [RAG Evaluation Workbench](<https://lucioduran.com/projects/rag-evaluation-workbench>) · [Source code](<https://github.com/lucioduran/rag-evaluation-workbench>)\n- [Bilingual Retrieval Evals](<https://lucioduran.com/projects/bilingual-retrieval-evals>) · [Source code](<https://github.com/lucioduran/bilingual-retrieval-evals>)\n\n[RAG: retrieval, search and evaluation](<https://lucioduran.com/topics/rag>)"},{"heading":"AI inference: measure latency, memory and quality","level":3,"url":"https://lucioduran.com/speaking#llm-inference","content":"**Audience:** Platform engineers, AI teams and people making infrastructure decisions.\n\n**What we explore**\n\n- Distinguish first-token wait, generation time and total duration.\n- Measure weight storage and process memory as separate variables.\n- Read a quantization experiment without extrapolating its results to another model or hardware.\n\n**Public reference material:**\n\n- [LLM Serving Bench](<https://lucioduran.com/projects/llm-serving-bench>) · [Source code](<https://github.com/lucioduran/llm-serving-bench>)\n- [Quantization Quality Lab](<https://lucioduran.com/projects/quantization-quality-lab>) · [Source code](<https://github.com/lucioduran/quantization-quality-lab>)\n\n[AI inference: latency, memory and quality](<https://lucioduran.com/topics/llm-inference>)"},{"heading":"Formats to discuss","level":2,"url":"https://lucioduran.com/speaking#formats","content":"Technical talk\n\nA presentation centered on a problem, the decisions that shape it and the available evidence, with time for questions.\n\nCode workshop\n\nA walkthrough of a public repository, its configuration and its tests. Scope depends on the group’s experience, time and available environment.\n\nSession for technology leaders\n\nA discussion of criteria for evaluating an AI proposal: what to test, which risks to control and which evidence to request before deciding."},{"heading":"How a session is prepared","level":2,"url":"https://lucioduran.com/speaking#preparation","content":"1. We agree on the audience, prior experience and the decision or skill the session should help develop.\n2. We select a topic and public materials that fit the scope. For a workshop, we define technical requirements and what will be run or reviewed.\n3. We distinguish observations from the experiments from claims that still need validation in another environment. References let participants continue working after the session."},{"heading":"Short introduction for organizers","level":2,"url":"https://lucioduran.com/speaking#speaker-bio","content":"Lucio Durán is C.T.O. at LOi and works from Uruguay. His public work combines software architecture with applied artificial intelligence, including open-source projects on agents, language model evaluation, RAG and inference. He publishes code, methods and results alongside each experiment’s scope and limitations.\n\n[Experience and education](<https://lucioduran.com/cv>) · [All projects](<https://lucioduran.com/projects>)"},{"heading":"Propose an invitation","level":2,"url":"https://lucioduran.com/speaking#invitation","content":"Send the organization or event name, audience, topic of interest, tentative dates, in-person or remote format and available time. With that context, we can discuss scope, requirements and terms.\n\n[Send a proposal](<mailto:email@lucioduran.com?subject=Invitation%3A%20artificial%20intelligence%20talk%20or%20training&body=Organization%20or%20event%3A%0AAudience%20and%20prior%20experience%3A%0ATopic%20of%20interest%3A%0ATentative%20dates%3A%0AFormat%20and%20location%3A%0AAvailable%20time%3A%0ASession%20objective%3A%0A>) · [email@lucioduran.com](<mailto:email@lucioduran.com>)\n\n[Other ways to get in touch](<https://lucioduran.com/contact>)"}],"links":[{"text":"Invite me to speak or propose a training session","url":"https://lucioduran.com/speaking#invitation"},{"text":"Agent Replay Runtime","url":"https://lucioduran.com/projects/agent-replay-runtime"},{"text":"Source code","url":"https://github.com/lucioduran/agent-replay-runtime"},{"text":"Agent Tool Capabilities","url":"https://lucioduran.com/projects/agent-tool-capabilities"},{"text":"Source code","url":"https://github.com/lucioduran/agent-tool-capabilities"},{"text":"AI agent architecture","url":"https://lucioduran.com/topics/ai-agents"},{"text":"Model Regression Gate","url":"https://lucioduran.com/projects/model-regression-gate"},{"text":"Source code","url":"https://github.com/lucioduran/model-regression-gate"},{"text":"Agent Injection Evals","url":"https://lucioduran.com/projects/agent-injection-evals"},{"text":"Source code","url":"https://github.com/lucioduran/agent-injection-evals"},{"text":"LLM and AI agent evaluation","url":"https://lucioduran.com/topics/llm-evaluation"},{"text":"RAG Evaluation Workbench","url":"https://lucioduran.com/projects/rag-evaluation-workbench"},{"text":"Source code","url":"https://github.com/lucioduran/rag-evaluation-workbench"},{"text":"Bilingual Retrieval Evals","url":"https://lucioduran.com/projects/bilingual-retrieval-evals"},{"text":"Source code","url":"https://github.com/lucioduran/bilingual-retrieval-evals"},{"text":"RAG: retrieval, search and evaluation","url":"https://lucioduran.com/topics/rag"},{"text":"LLM Serving Bench","url":"https://lucioduran.com/projects/llm-serving-bench"},{"text":"Source code","url":"https://github.com/lucioduran/llm-serving-bench"},{"text":"Quantization Quality Lab","url":"https://lucioduran.com/projects/quantization-quality-lab"},{"text":"Source code","url":"https://github.com/lucioduran/quantization-quality-lab"},{"text":"AI inference: latency, memory and quality","url":"https://lucioduran.com/topics/llm-inference"},{"text":"Experience and education","url":"https://lucioduran.com/cv"},{"text":"All projects","url":"https://lucioduran.com/projects"},{"text":"Send a proposal","url":"mailto:email@lucioduran.com?subject=Invitation%3A%20artificial%20intelligence%20talk%20or%20training&body=Organization%20or%20event%3A%0AAudience%20and%20prior%20experience%3A%0ATopic%20of%20interest%3A%0ATentative%20dates%3A%0AFormat%20and%20location%3A%0AAvailable%20time%3A%0ASession%20objective%3A%0A"},{"text":"email@lucioduran.com","url":"mailto:email@lucioduran.com"},{"text":"Other ways to get in touch","url":"https://lucioduran.com/contact"}],"structuredData":[{"@context":"https://schema.org","@graph":[{"@type":"Person","@id":"https://lucioduran.com/#person","name":"Lucio Durán","url":"https://lucioduran.com/"},{"@type":"WebSite","@id":"https://lucioduran.com/#website","url":"https://lucioduran.com/","name":"Lucio Durán","inLanguage":["en","es"],"publisher":{"@id":"https://lucioduran.com/#person"}},{"@type":"WebPage","@id":"https://lucioduran.com/speaking#page","url":"https://lucioduran.com/speaking","name":"Artificial intelligence talks and training","description":"Talk and workshop proposals by Lucio Durán in Uruguay: AI agents, LLM evaluation, RAG and inference, grounded in open-source projects and reproducible evidence.","inLanguage":"en","author":{"@id":"https://lucioduran.com/#person"},"publisher":{"@id":"https://lucioduran.com/#person"},"isPartOf":{"@id":"https://lucioduran.com/#website"},"isAccessibleForFree":true,"datePublished":"2026-11-10","dateModified":"2026-11-10","about":[{"@type":"Thing","name":"AI agents that act: permissions, failures and recovery","url":"https://lucioduran.com/topics/ai-agents"},{"@type":"Thing","name":"Evaluate models and agents before accepting an improvement","url":"https://lucioduran.com/topics/llm-evaluation"},{"@type":"Thing","name":"RAG with evidence: retrieval, languages and answers","url":"https://lucioduran.com/topics/rag"},{"@type":"Thing","name":"AI inference: measure latency, memory and quality","url":"https://lucioduran.com/topics/llm-inference"}]}]}],"contentHash":"69774841a290d868afa42d58959350eed03e81b4fce07e83985874a83606a890","provenance":{"method":"Deterministic serialization of the server-rendered main element","source":"https://lucioduran.com/speaking","hashAlgorithm":"sha256","hashScope":"JSON document fields before contentHash and provenance","sourceHTMLBytes":42901,"contentBytes":6055}}