# Chris Dabatos

> Staff DevRel Engineer at Fly.io. I build agentic workflows and AI-focused devtools, then turn what breaks into reference patterns developers can use.

Personal site: open source projects, technical deep-dives, and conference talks.
Every public blog post has a Markdown mirror at the same URL with `index.md` appended.
Posts also ship full audio narration, section by section.
An MCP server (Streamable HTTP) is available at https://chrisdabatos.com/mcp with tools: list_posts, search_posts, get_post, get_projects, get_talks, about.

## Writing

- [Don’t Build A Second Product For Voice](https://chrisdabatos.com/blog/dont-build-a-second-product-for-voice/index.md): A Deepgram Voice Agent pattern for adding realtime voice to an AI app that already has auth, memory, tools, and transcripts.
- [Your AI Agent Is Logged In. That Doesn't Mean It Should Have Access.](https://chrisdabatos.com/blog/ai-agent-access/index.md): What building RecallMEM taught me about agent authorization, vector search, delegated access, and why relevance is not permission.
- [Voice Agents Need Receipts](https://chrisdabatos.com/blog/voice-agents-need-receipts/index.md): RecallMEM now has a Deepgram voice agent with app-controlled memory. Voice made the memory layer stricter: no quote, no memory.
- [The LLM Was The Easy Part](https://chrisdabatos.com/blog/llm-was-the-easy-part/index.md): Building RecallMEM showed me the model call was easy. The real work was memory, files, tokens, setup, persistence, and product glue.
- [AI memory is broken because we let the model own the truth.](https://chrisdabatos.com/blog/ai-memory-is-broken/index.md): Building Speak2Me taught me why AI memory needs deterministic facts, transcript chunks, vector search, and app-owned truth.
- [AI engineering is web development all over again.](https://chrisdabatos.com/blog/ai-engineering-new-web-dev/index.md): The same cycle that made web development the path into tech is repeating with AI engineering, but the tools and learning curve are different.
- [Coding Used to Be Hard. That Was the Point.](https://chrisdabatos.com/blog/coding-used-to-be-hard/index.md): AI coding tools changed the craft. This is about what developers lose, what they gain, and why the hard parts still matter.
- [Voice agents should teach you, not just answer you.](https://chrisdabatos.com/blog/study-mode/index.md): I turned a voice-first AI journal into a Socratic tutor that quizzes through conversation and makes you explain the answer back.
- [Building an AI That Shows You Your Possible Futures](https://chrisdabatos.com/blog/parallel-lives/index.md): I built Parallel Lives, an AI app that turns decisions into branching timelines instead of another pros-and-cons list.
- [Voice apps fail when they treat speech like a text box.](https://chrisdabatos.com/blog/speak-it/index.md): I built a voice-to-text tool that learns writing style through statistics instead of storing private messages.
- [College rankings are useless if they ignore the money.](https://chrisdabatos.com/blog/college-picker/index.md): I built an AI tool that compares colleges by financial outcomes instead of prestige, rankings, campus photos, or vibes.
- [Journaling apps fail because they only remember what you typed.](https://chrisdabatos.com/blog/ai-journal-system/index.md): I built an AI journal that finds patterns, remembers context, and turns entries into a third-person life story instead of passive storage.

## Projects

- [Sprite Agent Workbench](https://github.com/RealChrisSean/sprite-agent-workbench): A dashboard for knowing which Sprite is awake, which checkpoint matters, and what state is safe to recover from.
- [RecallMEM](https://github.com/RealChrisSean/RecallMEM): An AI memory app that keeps chat, voice, files, facts, and recall in the same system instead of letting the model own the truth.
- [atlasMemory](https://github.com/RealChrisSean/atlasMemory): A unified AI memory layer with branching, rollbacks, and conflict detection — all powered by a single TiDB table.
- [tidb-ai-docs-copilot](https://github.com/RealChrisSean/tidb-ai-docs-copilot): Turns unstructured docs and GitHub issues into a structured TiDB knowledge graph using Bedrock embeddings + TiDB Vector.
- [semantic-qna](https://github.com/RealChrisSean/semantic-qna): A semantic FAQ assistant (CLI + web UI) using AWS Bedrock embeddings & TiDB Cloud vector search.
- [College Picker](https://github.com/RealChrisSean/college-picker): A college comparison tool that uses Department of Education data to make cost, debt, and outcome tradeoffs visible.

## Talks

- AI Memory Is Broken Because I Fixed It — Commit Your Code 2026, Sep 2, 2026: Every AI app claims it can "remember" you. However, most of them don't. So I started working on a fix. In this talk, I go over the three-layer memory architecture I built from scratch after the tools I trusted failed me. You'll see real failures, real fixes, and why your vector database choice matters more than you think.
- Persistent AI Memory Is A Systems Problem — DataConLA 2026, Aug 22, 2026: Persistent AI memory is not just retrieval. It is state, schemas, evidence, permissions, update rules, and recovery. I use RecallMEM to show why memory becomes a systems problem the moment it has to survive beyond a single chat.
- The LLM Was The Easy Part — AllThingsAI (Durham, NC), Mar 23, 2026: The model call was the easy part. The hard part was making memory believable: storing the right facts, keeping receipts, handling files, and knowing when retrieval is not permission. This talk walks through the failures that forced the system into shape.
- Meaningful Answers, Not Keyword Matches — OSA Con 2025, Nov 2025: Conference talk at OSA Con demonstrating how semantic search and AI can transform FAQ search from keyword matching to actually understanding what users are asking.
- Why Your FAQ Search Sucks and How I Fixed It — Tech Conference, Aug 20, 2025: Conference talk demonstrating how to improve FAQ search functionality using modern search techniques.
- Demystifying Data Replication with PostgreSQL — Percona Live, Jun 26, 2024: Conference talk at Percona Live exploring PostgreSQL replication strategies and best practices.
- SF Awesome AI Dev Tools — SF Meetup, Jun 16, 2025: Showcasing the latest AI development tools and frameworks at San Francisco meetup.
- From Vectors to Agentic Memory: Architecting for Supercharged AI Development — SF Meetup, Sep 23, 2025: Deep dive into AI architecture patterns and memory systems for modern AI applications.
- Open Source ELT For Everyone – Level Up With Custom Connectors — Big Data LDN, Oct 18, 2022: Conference talk at Big Data LDN covering open-source ELT pipelines and building custom data connectors.

## Optional

- [llms-full.txt](https://chrisdabatos.com/llms-full.txt): every public post in one file