Counselor: Turning a Rough Demo into a Working Prototype

Location

USA

Industry

EdTech

Duration

1 month

Team

2 Full Stack Engineers, 1 .NET Engineer

Type of service

Product Prototyping

Key technology

React, Python

  • 40K

    Active Users

  • 37

    Countries

  • 4M

    Monthly Recipe Views

About Client

Our client is Counselor, a US-based startup aiming to simplify the college decision process for teenagers. In the USA, it’s common for high school students to work with college consultants — professionals who help them choose the right college based on test scores, soft and hard skills, preferences, and long-term goals. Counselor aimed to turn this human-guided process into a digital experience. The app uses a detailed questionnaire and AI logic to generate personalized recommendations in a shortlist of three colleges that best match each student’s profile.

When the client contacted us, they already had an early version of the vibe-coded product. The frontend prototype was connected to an AI-based decision engine. The backend, however, was still in its early stages and prone to inconsistencies. Our job was to turn a rough demo into a working prototype, ready for demo and investor outreach.

The Challenge

Turning Concealer into a functional prototype was a challenging task, as the client came in with pre-built components, but the product lacked structure and predictability. Here is what we’ve faced:

  • Unpredictable AI responses. The AI often returned inconsistent outputs, forcing the team to test, tweak, and patch around it to ensure a stable user experience.

  • Limitations of pre-built tools. The “ready-to-go” solutions didn’t always behave as promised, requiring hands-on technical support to get them working properly.

  • Gaps between the idea and execution. The concept was strong, but the implementation was fragmented. We had to rebuild the vision into a prototype in the shortest term.

We got the chaos, applied our skills, and helped bring clarity and structure to the unstable prototype.

Goals

Stabilize Unpredictable Flows

Test and adjust the product logic to work around the AI’s inconsistencies and ensure reliable user interactions.

Improve Frontend Quality

Redesign the existing UI to create a visually appealing experience for teenagers and advisors.

Validate Prototype Potential

Turn a fragile prototype into a demo-ready tool that could be confidently shared with early users for feedback and iteration.

Dev Process

Frontend redesign

The client came to us with a half-working vibe-coded prototype — a visually not bad but chaotic front end. All of these were impossible to scale. Our first task was to bring structure and clarity. We rebuilt the UI with scalable code and design, making it user-friendly and visually consistent. Even though it wasn’t a full rebuild from scratch, the upgrade was noticeable and necessary to move the product beyond.


Testing and stabilizing unpredictable AI logic

We ran extensive tests across different user scenarios and consistently faced unpredictable results from the AI layer. Sometimes, the app would freeze or return irrelevant suggestions despite correct input. We added fallback logic and tweaks to ensure a more consistent workflow.


Wise use of already-made tools

This case showed the real-world limits of already-made platforms that claim to “do it all.” Our role was to make them behave. We optimized, stabilized, and made them provide predictable patterns. It wasn’t about building perfect code, but about smart fixes and creative problem-solving. This allowed the client to test their product idea quickly, without overspending on custom dev from day one.

Schema of the data migration process

Got a Project? Let’s Talk!

Design Process

We Started With Code

Our goal was to build a compelling product story that removed users from the idea of paying rent or managing a property and instead, transported them to a place where these tasks were actually exciting. How do we make paying rent rewarding? How do we create an experience where finding a tenant rewarding or closing on a broker commission is rewarding beyond the financial gain but on an emotional level?


We Started With Code

Our goal was to build a compelling product story that removed users from the idea of paying rent or managing a property and instead, transported them to a place where these tasks were actually exciting. How do we make paying rent rewarding? How do we create an experience where finding a tenant rewarding or closing on a broker commission is rewarding beyond the financial gain but on an emotional level?


We Started With Code

Our goal was to build a compelling product story that removed users from the idea of paying rent or managing a property and instead, transported them to a place where these tasks were actually exciting. How do we make paying rent rewarding? How do we create an experience where finding a tenant rewarding or closing on a broker commission is rewarding beyond the financial gain but on an emotional level?

Schema of the data migration process

  • 100K

    Conference Attendees

  • +25K

    Video Lectures

  • 150K

    Contributing Authors

Outcome

The Counselor project reflects a common and increasingly popular path: testing a concept quickly using AI tools before involving developers. In this case, AI helped validate the core idea without overinvesting. There were technical issues, but that’s where our team stepped in. We stabilized the product and prepared it for the next stage: making sense of chaotic setups, fixing what breaks, and adding the structure needed to move forward. We believe this kind of flow is a smart way to start.

The recent years have shown us that we shouldn’t start from scratch every time. Sometimes, we should take what’s already brought AI and make it work.

Along the way, we dealt with unpredictable AI behavior and limitations. But we took it, tested it thoroughly, and made adjustments where needed. The result is a clear, structured prototype that reflects the product’s core vision and is ready for demo and investor outreach.

“Jellyfish’s real-time communication and ability to adapt to InsideOut’s existing team communication tools enabled us to coordinate teams across multiple projects and manage how we worked together most efficiently. Prioritization and delivery of support to the InsideOut’s internal teams within a tight deadline were critical elements of the InsideOut and Jellyfish partnership.”

Daniel de Nieuwe

Product Lead at InsideOut

Is this case in line with what you have planned to build?

Let’s schedule a quick call to explore how we can support your business objectives