StreamScope

Tech Venture Bootcamp

During IE's Tech Venture Bootcamp, we built StreamScope, a real-time Twitch chat intelligence platform designed to help content creators and brands understand their audiences at a deeper level. We were driven by a simple idea: live streaming generates enormous amounts of data, yet creators have almost no tools to make sense of it in real time. StreamScope changed that. We built a full-stack platform combining NLP-based chat analysis and real-time audio transcription, going from concept to working product in just a few weeks. The project placed 2nd at the Bootcamp. Following the Bootcamp, we have continued developing StreamScope, driven by the belief that there is a real opportunity here worth pursuing.The experience pushed us to grow not just as engineers, but as a team, learning to move fast, make decisions under pressure, and build something we were genuinely proud of.

Project Overview

What was your project about?

StreamScope is a real-time chat intelligence system designed for live streamers. It ingests live chat messages, classifies them into structured engagement categories, and surfaces the most relevant interactions through a dashboard and overlays. The system is built to operate with low latency, transforming high-velocity chat into structured, actionable signals.

In essence, it turns chaotic chat into focused interaction.

Purpose of the project

What was the main purpose behind choosing this project, and what real-world problem were you aiming to address?

Live streaming places increasing cognitive pressure on creators. As audiences grow, chat velocity rises to a point where streamers can no longer meaningfully follow conversations, answer questions, or maintain context. Engagement quality declines not because of lack of effort, but because of human limitations.

We chose this project to address that structural constraint. Rather than slowing chat down, we focused on restoring clarity and preserving meaningful interaction as creators scale.

Technical Details

What technical approaches and tools did you use, and why?

We built StreamScope on a Python and Flask backend with server-rendered HTML and vanilla JavaScript to maintain low latency and architectural clarity. The system runs in a production-ready environment using a Hetzner virtual machine and an Nginx reverse proxy.

For classification, we adopted a deterministic-first approach. Messages are tagged using structured heuristic logic, with FastText serving as a fallback for low-confidence cases. An optional LLM path is scaffolded behind a feature flag but remains non-blocking to preserve performance and reliability. This layered architecture reflects our emphasis on stability, control, and real-time responsiveness.

What knowledge, tools, or ideas from your program influenced the project?

The keynote session significantly shaped our approach. Javier Pérez Trigo, Head of Digital Natives at Google, spoke about how early-stage teams succeed by prioritizing speed and focus over unnecessary complexity. That message resonated with us early on. By the end of the second day, we had moved past uncertainty and were energized, aligned, and ready to take on the challenge together.

After meeting with our mentor Aleix Catalan, we refined that focus further. He encouraged us to apply an impact–effort matrix to evaluate feature priorities. This framework helped us eliminate distractions, concentrate on high-value functionality, and avoid over-engineering. It gave structure to our decisions and reinforced disciplined execution.

We also applied principles from our coursework on modular architecture, reproducible environments, and production-oriented development, which directly influenced our backend design and deployment strategy.

Challenges and Solutions

Were there significant challenges during the experience?

One of the biggest challenges was maintaining scope discipline. It was tempting to expand into additional features such as collaboration tools or advanced analytics. However, we chose to stabilize and refine the core system before adding complexity.

Another challenge was balancing deterministic classification with AI fallback while preserving real-time performance. By structuring our architecture carefully and testing iteratively, we ensured the system remained stable and extensible.

Collaboration and Teamwork

How did teamwork contribute to the success of your project?

From the very beginning, our team dynamic played a major role in how quickly we progressed. We were three second-year Computer Science and Artificial Intelligence students, one first-year Data and Business Analytics student, and one master’s student in Computer Science and Business Technology. Each of us brought something different to the table.

Some of us had hands-on experience building production-ready mobile and web applications, working with real-time databases, authentication systems, and scalable backend services. Others had experience in AI systems, LLM-driven tools, and full-stack development. We also had strong exposure to product thinking, partnerships, and digital growth strategy. Charly brought the original idea and the analytical lens behind it, while Ines helped us stay organized and aligned from both a technical and strategic perspective.

Because of that mix, we were able to divide responsibilities clearly, move fast without stepping on each other’s work, and constantly challenge our assumptions. The balance between engineering depth and business clarity helped us stay focused and execute effectively under pressure.

Learning and Takeaways

What key lessons and skills did you strengthen?

Technically, we strengthened our skills in real-time system design, structured classification, deployment workflows, and architectural discipline.

From a soft skills perspective, we deepened our ability to prioritize effectively, communicate clearly under pressure, and align as a team around execution rather than abstraction.

Future Development

Do you plan to continue developing the project?

Yes. Our next step is structured validation with active streamers through pilot programs and questionnaires. We aim to improve classification accuracy, refine engagement analytics, and integrate more deeply into streamer workflows.

Over time, the structured chat data may enable deeper audience insights and cross-platform capabilities.

What advice would you give to other students?

If you are eager to participate and willing to put in the time and effort, the Tech Venture Bootcamp offers an opportunity to learn far beyond the classroom. It challenges you to think critically, move quickly, and build under real constraints.

Start with a real problem. Stay focused. Use structured decision-making tools. Commit early as a team. Clarity and discipline will take you further than complexity.