SureCare AI

Product overview

SureCare AI is an online symptom-assessment application. Tovra’s founder integrated the OpenAI API into its Django backend on a short contract engagement, focusing on the reliability and safety of the AI request path.

Objective

Give users a calm, structured interaction where every AI response shown has been validated, and where slow or failed requests are handled gracefully instead of breaking the session.

Key user flow

1. 1

   Start a session

   The user begins a guided, conversational assessment.

2. 2

   Answer structured questions

   The interaction collects symptom information step by step.

3. 3

   Continue safely

   The session is persisted, so a refresh does not lose the conversation.

4. 4

   Review a summary

   The user receives a structured, readable response.

Engineering challenges

• Untrusted output

  Model responses can drift from the expected structure. In this context, showing malformed or unexpected output to a user is not acceptable.

• Slow dependency

  The AI provider sits on the critical path of the conversation. Unbounded waits or retries would leave users stuck.

• Fragile sessions

  A conversation that disappears on refresh is frustrating and erodes trust.

Solution & implementation

A prompt and response contract enforced as a schema, with invalid output rejected at the boundary; explicit timeouts, bounded retries and graceful degradation on the request path; and persisted sessions.

1. 01

   Schema-enforced contract

   The prompt and expected response are defined as a schema. Output that does not validate is rejected at the boundary rather than rendered.

2. 02

   Timeouts and bounded retries

   Every provider call has an explicit timeout and a bounded retry policy.

3. 03

   Graceful degradation

   When the provider cannot produce a valid response in time, the application degrades to a clear, safe state instead of failing silently.

4. 04

   Session persistence

   Sessions are persisted so a conversation survives a page refresh.

Architecture

1. InterfaceGuided assessmentConversational flow (not built by Tovra)
2. APIDjango backendSession and request handling
3. AIOpenAI APISchema-validated responses, timeouts, retries
4. DataPostgreSQL + RedisPersisted sessions
5. InfrastructureDocker on AWSContainerised deployment

Technologies

• Python
• Django
• PostgreSQL
• Redis
• OpenAI API
• Docker
• AWS

Founder’s role

Oluwakorede Erinle · Backend & AI integration (contract) · Apr 2025 – May 2025

• Integrated the OpenAI API into a Django backend powering the assessment flow.
• Defined the prompt and response contract as an enforced schema, rejecting invalid output at the boundary rather than rendering it.
• Built the request path around a slow third-party dependency with explicit timeouts, bounded retries and graceful degradation.
• Persisted sessions so a conversation survives a refresh.

Engineering considerations

• Validate at the boundary: nothing reaches the user unless it matches the contract.
• Bound every wait. A predictable fallback is better than an indefinite spinner.
• Health-adjacent products need careful language and design that never overstates what software can do.

SureCare AI is presented here as an engineering project. Nothing on this page describes clinical accuracy or diagnostic capability, and symptom-assessment software is not a substitute for professional medical care.