# How Thinqpoint built real AI systems without the fluff > How Thinqpoint, a healthcare analytics startup serving safety net organizations, went from brittle raw API calls to a production agentic platform on Agno — compressing idea-to-production to a single week and answering multi-step queries in under a minute. - Published: 2026-07-21 - Author: Cosette Cressler - Category: Customers - Canonical: https://www.agno.com/articles/how-thinqpoint-built-real-ai-systems-without-the-fluff - Markdown: https://www.agno.com/articles/how-thinqpoint-built-real-ai-systems-without-the-fluff.md ## How Thinqpoint built real AI systems without the fluff We spoke with Bogdan Rau, Founder and CEO of [Thinqpoint](https://thinqpoint.com/), about taking the company from an experiment with AI to a full agentic system. Thinqpoint is a healthcare analytics startup focused on helping safety net organizations maximize their impact serving underserved communities, unlocking expert insights for federally qualified health centers, critical access hospitals, and community-based organizations. Bogdan is an epidemiologist by training, not a software engineer, and he built the platform without an engineering team behind him. In this post, we’ll start with where his first prototypes hit a wall, then look at how he picked a framework, how the platform runs on Agno today, what changed once it was in production, and how he thinks about building with AI at all. ### When raw API calls hit a wall Bogdan saw AI as a way to close long-standing analytics gaps in community health. He built a data visualization platform and tested whether AI could interpret thematic maps — a core need for the work. The early experiments were promising, and they also exposed hard limits. A single map interpretation took roughly 30 seconds. Each additional step in the analysis added more complexity. Without a framework, even simple AI-powered functionality took real effort: every new feature meant stitching together custom code, managing context by hand, and rethinking how analysis steps fit together. _“We built the first feature using Node.js, no framework, and a Gemini endpoint. There was a lot of heartache to just get it to run because there weren't any of the utilities you get with a framework.”_ Progress was possible, but slow and fragile. And Bogdan realized the thing he needed wasn’t faster API calls. _“We weren’t looking for API endpoints. We were looking for a way to orchestrate the work. It’s not just about what context you put in, but also when. We needed a framework that let us insert domain expertise at the right moments across a complex workflow. We wouldn't have been able to do that with any of the tools we evaluated back then, and the last thing I would have wanted to do is build that from scratch.”_ ### Picking pragmatism over flashy features Bogdan evaluated several API approaches and established frameworks, including LangChain and the OpenAI Agents SDK. Some offered flexibility but introduced significant complexity and boilerplate. Others required committing to a single model vendor. Either way, he could see himself spending more time managing infrastructure and vendor dependencies than improving the product. The first filter was vendor lock-in. _“Early on we had to make a decision: can we afford to choose a platform that locks us into one vendor? The answer was quickly no. And if not that, then what do you choose out of the sea of frameworks that are just now coming up?”_ But flexibility without pragmatism still fell short — the platform had to enable production-ready work without unnecessary complexity. Agno checked every box: clear documentation, an active community, no vendor lock-in, and a focus on the fundamentals needed to build real systems. _“Agno became a really good candidate because it had the main things I was looking for: simple implementation, a community to help onboard, and pragmatism—not the fluffy language and cool features that sound great on paper but break down when you try to implement them.”_ What sealed it was speed to validation. Within about a day, Bogdan tested Thinqpoint’s core hypothesis: whether inserting domain-specific knowledge at strategic points in a workflow could meaningfully improve analytical accuracy. It wasn’t a finished product, but it proved the idea worked and that there was a path from experiment to production. ### How the platform runs on Agno Thinqpoint runs on Azure, using a mix of serverless functions and containerized services. It integrates structured data — demographic tables, community health metrics — alongside unstructured content like documents, reports, and thematic maps, all managed through a custom-built content management system. Agno is deployed as a containerized service and plays two roles: - **On the backend**, it orchestrates the end-to-end data analysis workflow, drawing from both structured and unstructured sources and handling everything from data validation and quality checks to generating analytical artifacts - **On the frontend**, it exposes the APIs that power the user-facing application, coordinating the underlying processes when a user submits a question and returning results in a clear, actionable format The goal was never to add a chatbot. It was to orchestrate complex, multi-step analysis in a way that was fast to iterate, reliable in production, and sustainable for a very small team. ### What Agno changed day to day #### Rapid experimentation and validation Agno’s simplicity let Bogdan move from idea to working prototype fast. _“With Agno, we can start an idea at the beginning of the week and have it into production by the end of the week with evals in place.”_ That speed wasn’t only about shipping — it was about learning. _“A lot of the features we ended up implementing started with ‘I wonder if this would work,’”_ Bogdan says. _“And that very quickly turns into, ‘OK, how do we integrate it into our workflows?’”_ The fast feedback loop let Thinqpoint validate ideas early and discard what didn’t work, without committing months of engineering effort upfront. #### Orchestration over endpoints Rather than relying on one-off model calls, Agno let Thinqpoint orchestrate entire analytical workflows: managing context, sequencing reasoning steps, validating data quality, and generating usable artifacts. _“We needed more than one-off model calls that pass data in and get answers out. We needed a framework for orchestrating context. Not just what information the model sees, but when it sees it. Agno gave us the ability to strategically introduce the right pieces of information at the right moments across our workflows, so now we’re able to find the right timing when a piece of information is useful to the model to actually make a correct interpretation.”_ The orchestration-first approach produced results that were more reliable, more explainable, and better aligned with real analytical work than simple prompt-and-response interactions. #### Built-in tools for testing and debugging AgentOS became a critical part of the development process. Its chat-based interface let Bogdan test agent behavior, refine guardrails, and evaluate tone, accuracy, and workflow logic without deploying to a full development environment. _“AgentOS is kind of the instant proof of concept. It’s been invaluable.”_ That gave him confidence features were production-ready before they reached users. _“By the time I deploy something to development, I don't need to worry about it. We've already figured it out in AgentOS.”_ #### Concepts that map to real work For a data scientist building an agentic system, Agno’s design felt intuitive — it mirrored how Bogdan already thought about orchestrating analytical work. _“The core concepts make sense—not just from a software development perspective, but from how you orchestrate work as an individual or as a leader. There's a natural mapping between what we do day-to-day as humans and what you can do in Agno.”_ #### A partnership, not a vendor relationship Bogdan expected documentation and a few answered questions. What he got was collaborative problem-solving. _“The answers weren't just ‘here's how to use this feature.’ They were ‘have you thought about using this?’ or ‘we can't do that now—let's add it to the roadmap.’ That was fairly unique in my experience.”_ _“I don't think another platform or another team would have gotten me this far.”_ #### Continuous improvement, built in As the platform matured, Bogdan noticed improvements that weren’t coming from his own code. Updates to Agno’s core — how agents process instructions, how context passes between agents, how workflows are structured — were quietly improving Thinqpoint’s performance without changes on his end. _“It's not just improvement in code, but improvement in the way language is used to drive a workflow. That's led to faster answers and changes in how agents reply. Some of that doesn't come from us—it comes from improvements in the core platform itself.”_ ### What changed in production #### Faster time to insight Early experiments took 30 seconds just to interpret a single map. Thinqpoint now handles multi-step queries — understanding organizational context, pulling relevant data, running analysis, and generating recommendations — in under a minute. _“Had I used the old system, you can imagine—30 seconds over 30 seconds over 30 seconds. We'd be in the many-minutes range. Now with Agno, we do it in under a minute.”_ Against manual analysis the gap is wider still. _“We measured what it would take for someone to manually get the data, think about what it means, and arrive at an answer. It's always above tenfold. At least 10 times as long as going into our platform and asking the question.”_ #### More reliable, context-aware outputs The real value comes from moving past raw data into interpretation. _“Somebody will come to our platform and say, ‘My organization serves these zip codes, and I’m trying to figure out where I should do more outreach for food insecurity.’ There’s a lot built into that question.”_ The goal isn’t faster answers so much as recommendations that reflect real-world constraints, organizational context, and lived experience in community health. Instead of returning data, the platform generates recommendations informed by organizational context, population characteristics, and domain knowledge. #### Scalable productivity without added headcount Without a large engineering team, Bogdan needed a framework that amplified his own capacity. Agno cut the custom glue code, repetitive infrastructure work, and manual testing, letting Thinqpoint scale capability without scaling complexity. _“Some of the other platforms that I kind of tried out... just made it seem like I was at the bottom of a mountain. And to get to the other side was going to be some major lift. Getting this far, this quick would not have been possible without Agno.”_ #### Confidence to build toward bigger use cases With orchestration, evaluation, and iteration baked in, Bogdan is expanding toward more advanced workflows, including end-to-end research projects and multimodal interfaces. _“Agno for me wasn't just Agno. It was the framework itself and it came with the community. It came with subject matter expertise that I didn't necessarily have. Over time, Agno has become more than just the components we use. With agents in place, it’s really the glue that holds our platform together.”_ ### How Bogdan thinks about AI For Bogdan, AI’s value isn’t in the technology. It’s in solving real problems — and he’s seen how easy it is to get distracted by what’s possible and lose sight of what matters. _“I always try to step back and ask, ‘What problem are we actually trying to solve?’ A lot of the time, the problem you think you’re solving isn’t the real one. Without that effort to understand things at their core, there’s a real risk of building things that don’t matter. And especially now, with AI making software development so much easier, it’s more important than ever to be sure you’re solving the right problem, so you don’t end up spending months building the wrong thing.”_ He’s wary of a moment where AI has made it trivially easy to build something, but not necessarily something useful. _“Anybody can get a front-end interface in under a day now. That makes it even more important to make sure you're solving for the right problem, so you don't end up spending months on things that don't ever get used.”_ That conviction is what he saw reflected in the Agno team from day one. _“Early on, I was impressed by a conversation I had with Ashpreet. I remember thinking, why is this guy reaching out to me? He's a startup founder, he's busy. But it quickly became clear that it reflected how the whole team works—focused on understanding the problem we were trying to solve, not just selling us another framework.”_ For anyone evaluating agentic frameworks, his advice is simple: just start, and lean on the community rather than going it alone. _“I've gone down rabbit holes that don't lead anywhere, and had I just asked a question, I would have gotten a much better solution much quicker. Get involved in the community. Exploit the collective mind of everybody using Agno. There are a lot of folks willing to share their expertise.”_ ### What’s next for Thinqpoint The focus is shifting from answering individual questions to powering full-scale research and strategic planning workflows. _“What excites me the most is our ability to now build on these foundations. For example, our new needs assessment capability delivers end-to-end, complex analyses about the health and health-related social needs of communities. What started out as a chat interface that answers ad-hoc questions is now an advanced intelligence platform that delivers complex insights in minutes.”_ Rather than relying on users to know exactly what to ask, future iterations will help determine which questions matter in the first place, run the necessary analyses, and deliver outputs usable for real decision-making. As Bogdan put it: _“Somebody might come in and say, ‘I need to better understand the communities that I have.’ There isn’t one question built in there. There are hundreds of questions.”_ Thinqpoint is also exploring ways to meet users where they already work, whether that means integrating with messaging platforms or experimenting with multimodal interfaces like voice and video. At the same time, Bogdan is preparing the platform for deeper use of sensitive data, including PHI and PII. Security was a factor in choosing Agno from the start: user data stays fully within Thinqpoint’s environment, which wouldn’t be possible with a framework that routes data through external systems. _“The choices we're making now in terms of the platform we use and how we implement it are going to be critical when we start onboarding PHI, PII, or other types of sensitive information.”_ ### How to build a similar system using Agno Bogdan’s path suggests one thing worth copying: before you automate a workflow, be sure it’s the workflow that matters, and let the framework handle orchestration so you can spend your time on the domain expertise only you have. To start building, work through the [quick start guide](https://docs.agno.com/get-started/quickstart), then browse the examples in the [Agno repo](https://github.com/agno-agi/agno).