Case Studies
In-depth look at real-world challenges, solutions, and measurable outcomes
Coordinating Two Parallel SaaS Platforms for a UK University Under GDPR Constraints
Client: UK Higher Education Client — PEN Group · Higher Education / EdTech
Two workstreams — the legacy PHP student-records system (SIMS) and a new companion React/Node student portal — had to ship on independent but interlocking timelines for the same institution, coordinated across a 9-person team split across Backend, Frontend, SQA, and IT Support, under strict GDPR data-residency and retention rules for UK student records. Status reporting had been ad hoc and the two workstreams had drifted out of sync, risking a missed enrollment-cycle deadline.
Multi-Tenant SaaS Architecture for a B2B Job Marketplace
Client: Product Build — SDT LTD · HR Technology / SaaS
Building a B2B job marketplace where recruiting-agency tenants required contractual data isolation. A shared-schema approach was ruled out early. The challenge was designing a database-per-tenant architecture that could scale to 200+ tenants while keeping migration orchestration and connection pooling manageable for a 23-person cross-functional team shipping on a fixed go-live date.
Shipping an AI Content Platform on Someone Else's Rate Limits
Client: Product Build — SDT LTD · Marketing Technology / AI
SocialPro AI needed to generate and schedule content across multiple social platforms using OpenAI's API (text via ChatGPT, images via DALL·E) while running concurrently with a separate 23-person job-marketplace engagement. The core risk was OpenAI rate limits and cost: naive per-request calls during traffic spikes would throttle the whole platform and blow the API budget.
Scaling a PHP ERP from 80 to 1,000+ Concurrent Users
Client: Enterprise Client — Onest Tech LLC · Enterprise Software / Business Management
The flagship ERP product was hitting a ceiling at ~80 concurrent users. Response times degraded sharply under load, and the instinct from stakeholders was to rewrite the entire system as microservices — a 12-month project. As Tech Lead, I pushed back and proposed a profiling-first approach before any architectural changes.
Getting Facial-Recognition Attendance to 99%+ Accuracy in Low Light
Client: Product Build — Onest Tech LLC · HR Technology
Early pilots of the facial-recognition attendance system had unacceptable false-reject rates at poorly lit office entrances and warehouse floors. Employees were manually overriding the system so often it wasn't saving any admin time — undermining the entire premise of the product.
Scaling a Multi-Vendor Marketplace to 500+ Vendors Without a Storefront Rewrite
Client: Product Build — Onest Tech LLC · Retail & E-commerce
As vendor count grew past a few dozen, the original server-rendered storefront's search and category pages started timing out under load, and vendors were complaining about slow payout visibility. A full rewrite was on the table but would have paused vendor onboarding for months.
CI/CD Pipeline Rollout Across a 36-Person Engineering Team
Client: Internal — Onest Tech LLC · Software Development / DevOps
A 36-developer team across 8 concurrent projects, including the flagship HRM platform, was deploying manually via SSH. Staging was perpetually broken, deployments took 45+ minutes, and post-deploy incidents were frequent. Two previous attempts to introduce automation had failed because they tried to change too much at once, causing developers to bypass the pipeline entirely.
Building a Personalized News Feed on ISR and a Redis Ranking Layer
Client: Personal Project · Media / Publishing
Wanted to explore how far Next.js's Incremental Static Regeneration could go for a content site with a fast-moving, personalized homepage — the two are normally in tension, since ISR favors mostly-static pages while personalization favors per-request rendering.
Adding Live Classes and Paid Courses to an LMS Without Breaking the Grading Engine
Client: Product Build — Onest Tech LLC · Education & Training
The existing Laravel LMS handled course content, enrollment, and auto-graded assessments for 5,000+ students, but the business wanted paid premium courses (needing dual payment providers for different regions) and live virtual classrooms — both of which risked destabilizing the existing grading and progress-tracking engine if bolted on carelessly.