For over five years Azati has been a continuous engineering partner on a production-grade digital learning platform serving 1M+ registered learners and 75K+ teachers across multiple countries. The work spans backend service development, LMS integrations, assignment and assessment engines, high-volume data migrations, and reliability engineering on a system that genuinely cannot afford downtime.
registered learners across global markets
active teachers managing classes across multiple countries
database records processed and maintained without downtime
The client is an EdTech software company serving a major international educational publisher, delivering English language learning content built around National Geographic and TED materials to schools, universities, language centers, and corporate training programs across the US, Asia, Latin America, and Europe. The platform operates 24/7 across time zones, runs on a database growing by millions of records every week, and serves active classes where downtime is not an option.
Azati joined the engineering team as a backend specialist and has been an active contributor for over five years, covering everything from feature development and LMS integrations to production incidents and large-scale data migrations.
The platform serves users across the US, Asia, Latin America, and Europe simultaneously. There is no quiet window to take things offline, and every release has to be designed from the start to run without interrupting active users:
Student responses, gradebook entries, progress records: it adds up fast at this scale. Making structural changes to that data without locking tables or degrading queries is a real engineering constraint:
The system is a set of bounded-context services plus integrations with external LMS platforms via LTI. When something goes wrong, finding the source quickly is as important as the fix itself:
The product roadmap doesn't stop for reliability improvements. New course series, new assignment types, new teacher tools: all shipped while the platform serves active classes around the clock:
The client needed a backend engineer who could join an existing, complex codebase and contribute meaningfully without a long onboarding tail. The platform was already live and serving real users, so the bar for production-readiness was set from day one. Over five years the scope has been broad and ongoing:
Joining a five-year-old production platform with billions of records, multiple services, and a live user base is a different challenge from greenfield development. The Azati engineer passed a technical interview with the client's CTO and was selected for the ability to work independently in a complex Ruby on Rails environment. No extended ramp-up, no hand-holding through the architecture.
The same Azati engineer has been on this project for 5+ years. Continuity matters enormously on a platform of this complexity. Accumulated knowledge of how the system behaves under load, where the historical edge cases are, which integrations have quirks: that kind of context can't be onboarded quickly. Azati's staff augmentation model made it possible to maintain that continuity over the full engagement.
The Azati engineer participated in backlog refinement sessions, stakeholder demos, and coordination with the customer success team handling issues from real schools and universities. This isn't a contractor relationship where tickets get picked up in isolation: it's genuine integration into the product development process, from planning through to post-release support.
When incidents happen on a platform serving 1 million registered learners, there's no time for process overhead. The Azati engineer worked through production incidents, analyzed telemetry in NewRelic and Sentry, prepared incident reports, and contributed to retrospectives. That kind of operational discipline is built through experience, not process documents.
Whether you need to extend an existing learning platform or build one from scratch, Azati brings the backend depth and domain knowledge to do it right.
Let's talk about your platformThe assignment and assessment system is the part of the platform teachers use most directly. This covers the full lifecycle: creating and distributing assignments, collecting and grading responses, building test banks, running Online Placement Tests, and surfacing results in the gradebook. Auto-graded and manually reviewed responses are both supported, and grading pipelines run asynchronously via Sidekiq and SQS so they don't block the main application.
Schools using Canvas, Moodle, or Blackboard don't want to manage a separate login or manually sync grades. The integration layer handles LTI launch flows, SSO, grade passback, enrollment sync via OneRoster, and content linking across multiple LMS platforms, each with their own LTI implementation quirks. API contracts stay stable across web and mobile client versions throughout.
Structural changes to a database with billions of records require a different approach than typical migrations. Multi-step patterns add new columns alongside old ones, backfill data in batches, switch over, and clean up: all without the application going offline. A dedicated PERF environment, a copy of production with obfuscated data, is used for load testing migrations via Apache JMeter before they touch real user data.
In a distributed service architecture, knowing where a problem is matters as much as knowing how to fix it. Centralized structured logging across CloudWatch and Sentry, distributed tracing with request ID propagation, and monitoring alerts surface problems early. When incidents happen, the process covers telemetry analysis, root cause identification, hotfix deployment, and structured retrospective that feeds findings back into future release controls.
The platform supports a full teacher workflow across assignments, grading, analytics, and progress tracking, as well as hybrid classroom and self-study scenarios. Content covers the full CEFR spectrum across multiple course series, targeting all four language skills through interactive digital textbooks, audio and video materials, and mobile learning. The gradebook and analytics features provide progress visualization and performance breakdowns built on top of a large and continuously updating dataset.
| Metric / area | State at engagement start | Current state |
|---|---|---|
| Active students on platform | Growing early-stage user base | 1M+ registered learners globally |
| Active teachers | Limited institutional adoption | 75K+ active teachers across multiple countries |
| Data migrations | Ad-hoc, high-risk operations | Zero-downtime multi-step patterns via PERF environment |
| Release process | Basic deployment pipeline | Go/no-go reports, multi-environment QA, incident retrospectives |
| LMS integrations | Limited external connectivity | Canvas, Moodle, Blackboard, and others via LTI/OneRoster |
| Platform engagement | Single content series | Multiple course series across ages and CEFR levels |
The platform operates under ISO/IEC 27001:2022 and SOC 2 Type II certifications, covering information security management across the full platform lifecycle. Infrastructure is monitored 24/7 by a dedicated support partner under a permanent support contract. Quarterly penetration testing validates the platform's security posture on a regular cadence. All data handling complies with the security requirements of institutional clients including schools and universities across the US, Europe, and Asia.
The Azati engineer works on a Time & Material basis, embedded directly in the client team rather than operating as an external contractor. This means participation in team ceremonies, direct access to product stakeholders, and shared ownership of the platform's technical quality.
The team works in Agile sprints with backlog refinement, sprint planning, and demos to stakeholders. The Azati engineer's involvement spans the full delivery cycle:
Over five years of continuous engineering, the platform grew to serve over 1 million registered learners and 75,000 active teachers across schools and universities in the US, Asia, Latin America, and Europe.
Migration patterns and deployment practices developed during the engagement allow structural changes to a multi-billion-record database without taking the platform offline or degrading performance for active users.
LTI integrations with Canvas, Moodle, Blackboard, and others give institutional clients the ability to deliver content within their existing LMS, with grade passback and enrollment sync handled automatically. This is a direct factor in adoption by schools and universities with standardized LMS infrastructure.
The release process that evolved over the engagement, multi-environment testing, go/no-go reports, PERF load testing, and structured incident retrospectives, means the team can ship confidently on a platform where failures affect real students in real classrooms.
Five years on a platform of this complexity leaves real expertise inside Azati: LMS/LTI/LRS/OneRoster standards, blended learning architecture, high-volume educational data patterns, and the operational realities of running a global 24/7 SaaS for educational institutions.
Some of what was built here goes beyond individual features: it's infrastructure and process that the platform now depends on:
Setting up a production-mirroring PERF environment with Apache JMeter load simulation changed how risky changes get evaluated. Instead of finding out in production that a migration takes six hours under real traffic, the team finds out in PERF first. That capability has probably prevented more incidents than any other single investment in the engagement.
The patterns developed for changing large production tables without locking the app, multi-step column changes, concurrent index builds, batched retroactive backfills, are now the standard approach. Not a workaround, but how the engineering team thinks about database changes by default.
Centralized structured logging, distributed tracing, and request ID propagation across services: the real payoff shows up during incidents. The ability to trace a user-reported problem through multiple services to the exact query or service call that failed is the difference between a 20-minute resolution and a 4-hour one.
LMS integration isn't just a technical feature, it's what makes the platform adoptable by universities and large school districts that have standardized on Canvas or Moodle. The LTI integration layer built and maintained over the engagement is directly tied to the platform's ability to serve institutional customers at scale.
The engagement has run lean on the Azati side by design: a single embedded engineer with deep platform knowledge, working as an integral part of the client team rather than as external support.
Custom Redmine plugin development and full-stack upgrade for a DevOps MSP, 30+ plugins adapted, Ruby 3 and Rails 7 migration, zero business disruption.
Mobile app with CNN model for automated steak marbling quality assessment.
Secure locally-hosted LLM alternative for confidential corporate communication.
LLM-powered microservice for automated candidate selection with semantic search.
Custom Jira Service Desk plugin with SLA calculation and task management features.
Front-end development for agriculture platform with improved UX and filtering.
Last updated