T.
Agentic AI Class Project · Para-Athletics Dashboard

Helping para-athletics coaches turn paper records into performance insight.

Project context

This was part of my AI Applications class, where the assignment was to develop an agentic AI application. The idea came from conversations with my sister, a Paralympic coach, about the realities of developing visually impaired athletes in Botswana and the amount of performance information still being kept in notebooks, WhatsApp messages, and individual memory. I then shaped the problem through conversations with coaches and athletes who had experienced those gaps from both sides of the track.

What stood out was how much coaching knowledge existed, but how little of it was captured in a usable system. Athlete profiles, training logs, event results, injury notes, and performance progress were still being tracked through fragmented paper records, making it difficult to monitor improvement, compare athletes, prepare for competitions, or build credible profiles for sponsorship and federation support.

My contribution

I owned the product discovery, product vision, and prototype build. I conducted user interviews, synthesized the coaching pain points, defined the dashboard structure, and began building a Streamlit-based prototype with AI-assisted query and insight features. I used GenAI as a technical partner while learning Python, local LLM setup, API calls, and app development workflows.

Problem

How might grassroots para-athletics coaches track athlete performance, training progress, and readiness when most records are still fragmented across notebooks, paper logs, and informal coaching memory?

Outcome

Designed and began building Lobelo, a lightweight AI-supported performance dashboard that helps coaches manage athlete profiles, training logs, event results, readiness signals, and performance insights in one place.

From paper logs to performance intelligence

The early prototype grew out of field interviews and real coaching artifacts. Coaches were already collecting valuable information, but the data lived in notebooks, WhatsApp updates, and individual memory. Lobelo is my attempt to turn that information into a structured tool coaches can actually use.

01

Phase 1: Field discovery with coaches and athletes

Conducted 6 problem-identification interviews with coaches, athletes, and volunteer sport leaders who had experienced the performance-development gap firsthand, including people working directly with visually impaired athletes in village schools.

6 problem-identification interviews · Coach and athlete discovery · Village-level sport development
02

Phase 2: Translating paper records into product requirements

Used examples of athlete profiles, training notes, event logs, and performance records to define the core data model, dashboard structure, and coaching workflows for tracking progress over time.

Athlete profiles · Training logs · Event results · Readiness tracking
03

Phase 3: Prototype build and usability testing

Built an early Streamlit prototype and ran 4 usability tests to understand whether coaches could navigate athlete profiles, training logs, readiness signals, and AI-assisted query flows.

4 usability tests · Python · Streamlit · Local LLM · Agentic AI workflow
Prototype features

What Lobelo is being designed to support

01

Athlete profiles

Centralizes athlete details, event category, impairment class, contact context, training needs, and performance history.

02

Training and event logs

Turns paper-based session notes and competition records into structured data that can be tracked over time.

03

Readiness and performance insight

Helps coaches monitor consistency, injuries, progress, and event readiness across athletes and competitions.

04

Ask Lobelo assistant

Allows coaches to query athlete data in plain language and receive summaries, comparisons, and coaching-relevant insights.

Building an AI product from field discovery

I did not approach this as a traditional software project. I started by listening to coaches, athletes, and volunteers who were already doing the work with limited tools. From there, I used GenAI as a build partner while learning to run Python workflows, structure CSV data, build Streamlit screens, test local LLMs, and connect external APIs.

The technical challenge was not only building screens. It was figuring out how to turn messy, real-world coaching information into a product system: what data to capture, how coaches would search it, what insights would be useful, and how an AI assistant could support decision-making without replacing the coach’s judgment.

Takeaway

This project taught me that AI product design starts before the model. The most important work was understanding the environment: volunteer coaches, limited infrastructure, paper-based records, visually impaired athletes, and the need for a lightweight tool that could fit into real coaching routines. Lobelo helped me see how AI can support overlooked systems when the product is grounded in the people already doing the work.

From notebooks to a coaching dashboard

Early discovery showed that coaches were already collecting useful athlete and training data, but it was difficult to preserve, compare, or act on. Lobelo brings those records into a structured dashboard designed for coaching decisions.

Handwritten para-athlete training and event log used before Lobelo
Before: paper coaching records
Lobelo para-athletics performance management dashboard prototype
After: Lobelo dashboard prototype