Uppr: Real-Time Meeting Assistant
Founding Engineer & Director. From Antler proof of concept to production SaaS.
Built the frontend and real-time SignalR meeting experience for Uppr, surfacing relevant Confluence and Guru notes during live calls. Took the product from an Antler-funded PoC to sub-5-minute deploys and an analytics-led pivot.
"How do we handle SOC2 and security questionnaires during enterprise pilots?"
Problem
Before modern LLM copilots were ubiquitous, Uppr set out to solve a focused problem for customer-facing teams: surfacing the right internal notes from Confluence, Guru, and Notion in real time while a live meeting was happening, without forcing the user to search across browser tabs mid-conversation.
As Founding Engineer and one of the company's directors, I collaborated with two co-founders to build the proof of concept we pitched to Antler to secure funding, and then owned the frontend architecture, real-time meeting experience, and cloud observability as we scaled to pilot customers.
Architecture
Uppr combined a Next.js + TypeScript frontend with a .NET / C# REST & SignalR backend backed by SQL Server and Azure Blob Storage:
- Real-Time Meeting Stream (
SignalR+Azure Speech SDK+Zoom Apps SDK): Browser microphone capture or a Zoom/Recall meeting bot streamed live transcript context toNoteHub.cs, pushing partial keyword matches and retrieved note cards back to the client over WebSockets/SignalR. - Knowledge Integrations & Bulk Onboarding: Connected team workspaces to Confluence and Guru (with Notion and Salesforce models in progress), supported by rich-text note editing (
TipTap+Mantine) and bulk import flows. - Analytics & Funnel Reporting (
pages/reporting.tsx): Built a real-time reporting dashboard aggregating group and user session metrics, giving the founding team visibility into where users dropped off between signup and live meeting usage. - End-to-End Cloud, Auth & Security: Managed Auth0 authentication, GitHub Actions CI/CD, Sentry error tracking across client and edge routes, and Playwright +
@axe-core/playwrightautomated E2E and accessibility tests.
// Real security headers middleware hardened for our CREST-accredited pen test
public Task InvokeAsync(HttpContext context, RequestDelegate next)
{
if (!_isDevelopment)
{
context.Response.Headers.StrictTransportSecurity = "max-age=31536000";
}
context.Response.Headers.ContentSecurityPolicy = _contentSecurityPolicy;
context.Response.Headers.XFrameOptions = "DENY";
context.Response.Headers.XContentTypeOptions = "nosniff";
context.Response.Headers.Add("Cross-Origin-Embedder-Policy", "require-corp");
context.Response.Headers.Add("Cross-Origin-Resource-Policy", "same-origin");
return next(context);
}Key Trade-offs
| Decision Axis | Chosen Path | Rejected Alternative | Why It Matters |
|---|---|---|---|
| Live Meeting Transport | Microsoft SignalR Hub (NoteHub) | Short-interval HTTP polling | Delivered sub-second partial matches and live transcript updates natively alongside our .NET backend. |
| Frontend Delivery Velocity | Next.js + Mantine + SWR + Playwright | Heavy custom UI kit from scratch | Allowed a sole frontend engineer to ship accessible forms, modals, and real-time views with under-5-minute deploys. |
| Enterprise Readiness | Strict CSP/COOP/HSTS middleware + CREST pen test | Deferring security hardening post-seed | Unblocked enterprise security reviews early by locking down headers, Auth0 flows, and DOM sanitisation (DOMPurify). |
Outcome
- Pitched the live working proof of concept to Antler with two co-founders and secured startup funding.
- Maintained a sub-5-minute deployment pipeline as the sole frontend developer while shipping Zoom App support, live SignalR note retrieval, and reporting dashboards.
- Identified key user-journey churn points via real-time analytics, driving a data-backed product pivot.
- Completed security remediation following a penetration test by a CREST-accredited organisation.