# Xlinfe > Xlinfe is an AI studio. We build products and systems that help people find answers, complete tasks, and serve customers. This site documents public work, services, and technical writing. ## Agent guidance Use the canonical URLs below when you cite the site. Read the full Markdown representation at `https://www.xlinfe.com/markdown/{page-id}` when you need the page body. Do not infer private client, personal, or unpublished information. ## Public profile - Website: [https://www.xlinfe.com](https://www.xlinfe.com) - Contact: [Contact form](https://www.xlinfe.com/#contact) ## Pages - [Home](https://www.xlinfe.com): Xlinfe is an AI studio. We build products and systems that help people find answers, complete tasks, and serve customers. - [Services](https://www.xlinfe.com/services): AI development, automation, web and mobile apps, and 3D experiences. Explore our services, data tools, technology choices, and ways to work with Xlinfe. - [3D Walkthroughs for Pre-construction Sales](https://www.xlinfe.com/3d-walkthroughs): Buyers walk through your project at eye level, from a link on their phone, tablet, or laptop. We build it from your plans, with a panel that shows which rooms each buyer visits. - [Work](https://www.xlinfe.com/work): Explore selected projects from Xlinfe. See what we built, the problems we addressed, and how each system works. - [Knowledge](https://www.xlinfe.com/knowledge): Practical notes from Xlinfe on AI models, tools, experiments, and engineering. Explore evaluations, voice agents, source retrieval, and browser 3D. - [About](https://www.xlinfe.com/about): Xlinfe is an applied AI studio based in the Dominican Republic. Work directly with founder Erik Mejia, from defining your project to building and launching it. - [Book a call](https://www.xlinfe.com/schedule): Book a 60-minute call with Xlinfe to discuss what you want to build or improve, how we can help, and the next step. - [AI development tools](https://www.xlinfe.com/stack): Tools and practical guides for building, testing, and operating AI products. - [Cursor guide](https://www.xlinfe.com/stack/cursor): Practical workflows and review checks for Cursor. - [Claude Code guide](https://www.xlinfe.com/stack/claude-code): Practical workflows and review checks for Claude Code. - [v0 guide](https://www.xlinfe.com/stack/v0): Practical workflows and review checks for v0. - [Lovable guide](https://www.xlinfe.com/stack/lovable): Practical workflows and review checks for Lovable. - [How to compare AI models](https://www.xlinfe.com/ai/models): Compare models on task quality, cost, response time, and data requirements. - [Sitemap](https://www.xlinfe.com/sitemap.xml): Every public page, as XML. ## Case studies and notes - [Flowcost](https://www.xlinfe.com/work/flowcost): A visual planning tool for AI workflows. Teams map agents, tools, and data, compare model and infrastructure choices, and estimate cost and response time before building. - [NVIDIA](https://www.xlinfe.com/work/nvidia): An internal hub where NVIDIA teams share and discover AI projects, agents, and workflows. Submissions go through review before publication. Built with Candid Leap for NVIDIA. - [SASD](https://www.xlinfe.com/work/sasd): A 3D walkthrough of a proposed assembly hall, built from architectural plans. Visitors explore in their browser, ask a voice guide questions, and open plans as they walk. - [Hallbase](https://www.xlinfe.com/work/hallbase): Congregation software for reports and territories. An AI assistant on WhatsApp or Telegram lets members check assignments, submit reports, and access permitted records. - [Graphite](https://www.xlinfe.com/work/graphite): A Webflow app for finding technical SEO issues, checking site health, and updating titles and descriptions across multiple pages. Built with Finsweet for Graphite. - [Candid Leap](https://www.xlinfe.com/work/candid-leap): AI development training and software improvements for Candid Leap’s team. We helped them choose models, test results, and improve code, security, and releases. - [Guessback](https://www.xlinfe.com/work/guessback): A voice game where players describe a word and an AI tries to guess it before time runs out. It tracks completed words, shows round results, and lets friends share challenges. - [Didastik](https://www.xlinfe.com/work/didastik): A learning platform that turns study material into practice questions and a weekly study plan. It schedules the next review using quiz results, past reviews, and study priorities. - [Turning architectural plans into a browser walkthrough](https://www.xlinfe.com/knowledge/plans-to-a-browser-walkthrough): Lessons from a browser 3D walkthrough: a rebuildable Blender model, baked lighting, streamed assets, performance checks, and a guide using plan sources. - [A voice agent’s rules can’t live in its prompt](https://www.xlinfe.com/knowledge/voice-agents-need-hard-limits): Use source tools, server-side action checks, and session limits to control voice agents. Prompts alone cannot enforce permissions or spending. - [Context engineering for coding agents](https://www.xlinfe.com/knowledge/context-engineering-for-coding-agents): Give coding agents clear repository instructions, working checks, and useful examples. Make each task easier to implement and review. - [Choosing a model is an eval problem](https://www.xlinfe.com/knowledge/choosing-a-model-is-an-eval-problem): Compare AI models on your own tasks. Measure answer quality, cost per task, and response time, then test routing and fallback options. - [Evals start with real tasks](https://www.xlinfe.com/knowledge/evals-start-with-real-tasks): Build AI evaluations from real tasks, define expected outcomes, and test critical failures. Use the results to guide release and ongoing review. - [A grounded answer shows its source](https://www.xlinfe.com/knowledge/grounded-answers-need-sources): Build knowledge agents with traceable sources, version details, access controls, and clear responses when information is missing or conflicts. - [Self-improving agents need evidence loops](https://www.xlinfe.com/knowledge/self-improving-agents-need-evidence-loops): Turn expert corrections into reviewed evidence, targeted evaluations, and software changes. A practical approach to improving AI agents after launch. - [Sandboxing agents requires real enforcement](https://www.xlinfe.com/knowledge/sandboxing-agents-requires-real-enforcement): How to assess a coding-agent sandbox: file access, network rules, child processes, and the actions that still need approval. - [Voice AI latency depends on more than the model](https://www.xlinfe.com/knowledge/voice-ai-latency-is-a-routing-problem): How network routing, session setup, and interruption handling affect voice AI. Engineering lessons from OpenAI’s real-time media infrastructure.