10x
concurrent learners supported
99.95%
uptime through exam weeks
-45%
infrastructure cost per learner
3 weeks
to staff a 6-engineer pod
The Client
An EdTech company operating a learning management platform used by universities and corporate training programs. After landing several large institutional contracts in a single year, enrollment was on track to grow roughly tenfold — on an architecture originally built for a fraction of that load, maintained by a small in-house team fully occupied with feature commitments.
The Challenge
The warning signs were already visible: dashboard pages timing out during peak evening hours, video lectures buffering for remote cohorts, and one partial outage during a client's midterm week that nearly cost the contract. The platform ran on a handful of oversized servers, every course page hit the database directly, and video was served from the application's own storage.
The in-house team knew what needed to change but had no spare capacity — and hiring senior platform engineers locally was a six-month proposition. The next enrollment wave, including hard exam-week deadlines, was two semesters away. They needed senior engineers embedded in their team within weeks, not quarters.
What We Did
Dillo staffed a dedicated 6-engineer pod — four backend/platform engineers, one DevOps engineer and one performance-focused QA — in three weeks from the first call. The pod worked in the client's Slack, GitHub and sprint cadence, with the client's CTO owning priorities and Dillo owning employment and retention.
- Profiled and rewrote the heaviest PostgreSQL query paths — course dashboards, gradebooks, activity feeds — eliminating N+1 patterns and adding the right indexes, cutting p95 latency on key pages dramatically.
- Introduced a Redis caching layer for session data, course structures and computed progress, taking the bulk of read traffic off the primary database.
- Moved the platform to autoscaling groups on AWS, provisioned with Terraform, replacing hand-managed servers with infrastructure that scales for evening peaks and idles overnight.
- Shifted video delivery to CDN-backed streaming with adaptive bitrates, fixing the buffering problems for remote and international learners.
- Built a load-testing harness simulating exam-week traffic shapes — thousands of simultaneous quiz submissions — and made passing it a release gate.
Because this was staff augmentation rather than a handoff project, every architectural decision was made with the in-house team in the room. Pricing stayed the same throughout: transparent salaries plus Dillo's low flat margin on every invoice.
The Results
- The platform now supports 10x the concurrent learners it handled at the start of the engagement, verified continuously by the load-testing harness.
- 99.95% uptime through exam weeks — the highest-stakes traffic windows of the academic year — with no repeat of the midterm-week incident.
- Autoscaling and caching cut infrastructure cost per learner by 45%, even as absolute traffic multiplied.
- The 6-engineer pod was fully staffed in 3 weeks, versus the ~6 months the client had budgeted for equivalent local hires.
- The client renewed the pod after the re-architecture and pointed it at the next challenge: real-time collaboration features on the now-stable platform.
Figures reflect outcomes reported for this engagement; they are project results, not audited benchmarks.
Tech Stack
Services Used
IT Staff Augmentation — a dedicated platform-engineering pod embedded in the client's team, on Dillo's transparent flat-margin pricing.