Case StudiesInfiniteWatch: High‑Throughput Event Ingestion
High‑throughput event ingestion for an AI‑native customer interaction intelligence platform
InfiniteWatch, an AI-native customer interaction intelligence platform, partnered with Fibertide to redesign the core event ingestion pipeline. The result: a scalable, production‑grade architecture ready for enterprise growth.
Collaboration Overview
The Client
InfiniteWatch is an AI‑native customer interaction intelligence platform. Its AI agents monitor customer interactions across web sessions, voice channels, and support channels, automatically detecting issues, and surfacing actionable insights rather than raw metrics. Think analytics, but with the human interpretation already done.
InfiniteWatch has raised $4M led by Base10 Partners, with scout investors from Sequoia and a16z, and was founded by a team of repeat founders with a successful prior exit.
Why did they need us?
InfiniteWatch's product depends on ingesting high volumes of session events in real time, with demanding latency, throughput, and availability requirements.
As the platform matured and larger customers came into view, the ingestion architecture needed to evolve, from its initial implementation to a system designed for enterprise‑grade throughput, reliability, and observability.
The team needed a partner with deep infrastructure expertise to execute this transformation while they stayed focused on their core AI product.
How did we help?
We ran the ingestion rebuild as a dedicated workstream alongside InfiniteWatch's engineers, taking end‑to‑end ownership of the infrastructure work. Phased, milestone‑driven delivery kept the client in control at every step, letting their team stay focused on the AI product.
We helped redesign InfiniteWatch's ingestion pipeline into a modern, event‑driven architecture, cutting memory usage on the new ingestion proxy by 12x and dropping p95 latency from 1,050 ms to 56 ms, giving them the reliability and scalability their AI‑powered product demands.
Client Insights
Company Type:
Startup
Location:
NYC, US & Madrid, Spain
Industry:
AI, SaaS, Customer Intelligence
Technology:
Kafka, ClickHouse, Go, MongoDB, Datadog, Confluent
"Fibertide takes end‑to‑end ownership of deep infrastructure work. From day one there was no doubt we were working with real experts in the field. Communication was excellent: we had daily updates, feedback was accommodated fast, and key milestones landed on or ahead of schedule. They rebuilt the backbone of our platform while my team stayed focused on what actually makes InfiniteWatch unique."
Alberto González Lizán
CTO & Co‑Founder at InfiniteWatch
The Challenge
Scale ingestion for growing traffic and larger customers
Enable horizontal scaling without re‑engineering
Decouple ingestion from downstream processing
Establish production‑grade observability and CI/CD
Our Solution
Stateless ingestion proxy with auth and rate‑limiting
Async processing via dedicated Kafka consumers
OpenTelemetry tracing, logging, and metrics
Automated deployments from day one
Results
Clean, event‑driven pipeline with separation of concerns
Dramatically leaner services, with p95 latency under 100 ms at load
Full observability from the first deployment
Milestones delivered on time, phase by phase
Project Highlights
Phased, Transparent Delivery
We structured the project into clearly defined phases, from proxy core through authentication and traffic control to asynchronous processing, each with visible milestones and deliverables.
This approach gave InfiniteWatch full visibility into progress and allowed them to validate each layer before moving to the next. Delivery stayed on track throughout, building confidence in the partnership.
Observability & Deployment from Day One
Rather than treating observability and CI/CD as afterthoughts, we embedded them from the start. OpenTelemetry‑based tracing, structured logging, and metrics were configured alongside the first deployable service. Automated deployments to test environments gave the team a reliable feedback loop for every change.
Forward‑Looking Architecture Review
Beyond the ingestion rebuild, we reviewed InfiniteWatch's insight‑generation pipeline and mapped out concrete reliability and efficiency improvements for the next phase of growth, so when the team is ready to tackle that work, the roadmap is already there.
High‑Performance Ingestion Architecture
The heart of the engagement was a full redesign of InfiniteWatch's event ingestion system. The impact was immediate and measurable. The new ingestion proxy's:
- memory usage dropped 12x;
- p95 latency fell from 1,050 ms to 56 ms;
- and p99 latency fell from 1,460 ms to 99 ms.
We built a new lightweight, stateless proxy whose sole responsibility is to authenticate requests, enforce rate limits, and forward events to Kafka, while all downstream processing (session creation, metadata enrichment, analytics) now happens asynchronously through dedicated consumers.
Key Outcomes
The rebuild gave InfiniteWatch the infrastructure foundation to pursue enterprise customers with confidence, while keeping their team focused on the AI product.
Startup‑Matched Velocity
Our delivery cadence matched InfiniteWatch's startup pace, with key milestones landing on or ahead of schedule so each layer could be validated before the next was built on top.
Operational Maturity
Day‑one observability and automated deployments shortened feedback loops, reduced on‑call risk, and set the team up to ship confidently as the system grows.
Enterprise Readiness
The ingestion pipeline is now production‑grade, ready to stand up to enterprise customer requirements and investor technical due diligence.
Scalable by Design
An event‑driven architecture with headroom to spare, built to absorb traffic spikes and support enterprise‑scale customers without re‑architecture.
Ready to build your success story?
Let's discuss how our expertise can make your product market-ready.
Get in touch