Work that shipped
Real projects. Real outcomes. Here's what we've built — and what it meant for the businesses behind them.
FinTech · Ongoing partnership
MarketAlerts.ai
AI-Powered Investment Intelligence
MarketAlerts.ai needed a system that could ingest, process, and classify massive volumes of financial data in real time — news feeds, SEC filings, earnings reports, social sentiment — and turn it into actionable signals for retail and institutional investors.
We designed and built the full pipeline: data ingestion from 20+ sources, LLM-powered signal classification with human-calibrated confidence scoring, a real-time alerting system with per-user filtering, and a dashboard for monitoring and backtesting signal accuracy.
The platform processes thousands of data points per hour with 95% classification accuracy, and has been running in production continuously since launch.
The Challenge
Financial data is noisy, time-sensitive, and high-stakes. The client needed a system that could separate meaningful signals from noise — fast enough to be actionable, accurate enough to be trusted.
Our Approach
We started with a focused MVP: one market, one signal type, one alert channel. Validated accuracy against human analysts, then expanded market coverage and signal types incrementally. Each expansion was measured before the next.
Tech Stack
EdTech · 8+ years (ongoing)
ClassCharts
Scaled to Millions of Users
ClassCharts is the leading behaviour management and seating plan tool used by thousands of schools across the UK. We joined as the core development team when the product was in early stage, and helped scale it to serve millions of students, teachers, and parents.
Over an 8+ year partnership, we built and maintained the full stack: the web app used by teachers and school administrators, the parent and student mobile apps, the data pipeline handling millions of daily events, and the infrastructure that keeps it all running at 99.9% uptime.
The work spanned everything from feature development to cloud migration, from performance optimization to building the mobile apps from scratch. We embedded as the engineering team — not an outside vendor, but the people who shipped the product.
The Challenge
School software has extreme usage spikes (morning registration, end of day, report card season) and zero tolerance for downtime. The platform needed to scale reliably while shipping features fast enough to stay ahead of competitors.
Our Approach
Deep integration with the founding team. We operated as their engineering department — attending standups, owning the roadmap, making architecture decisions. This wasn't outsourcing; it was a partnership.
Tech Stack
Government / Smart City · 18 months
Smart City IoT Platform
Government Infrastructure
A Middle Eastern government contracted a consortium to build the digital infrastructure for a new smart city development. We were brought in to design and implement the backend architecture — the layer connecting physical IoT sensors to the applications that city operators and residents would use.
The platform spans 5+ city domains: smart parking (sensor-driven occupancy and guidance), electricity grid monitoring, water distribution tracking, waste collection optimization, and environmental monitoring. Each domain has its own microservices, data pipelines, and operator dashboards.
We built 30+ microservices on Kubernetes, designed the event-driven architecture connecting IoT gateways to application services, and implemented the data layer handling millions of sensor readings per day.
The Challenge
Government infrastructure projects have strict requirements: multi-vendor coordination, long timelines, formal documentation, and systems that must run for decades. The architecture needed to be modular enough for independent domain teams to build on.
Our Approach
Domain-driven design from day one. Each city domain (parking, electricity, water, etc.) got its own bounded context with clean API contracts. This let multiple vendor teams build in parallel without stepping on each other.
Tech Stack
EdTech · 8+ years
ProvisionMap
Nationwide School Provision Tracking
ProvisionMap helps schools track and manage student support provisions — from one-on-one tutoring to group interventions — and measure whether those provisions are actually working. It's used by thousands of schools across the UK for Ofsted reporting and SEND compliance.
We led the full engineering effort over an 8-year partnership: building the core mapping engine, the reporting dashboard used by SENCOs and headteachers, and the data export pipeline that feeds into government compliance systems.
The platform grew alongside ClassCharts, sharing infrastructure while maintaining its own domain-specific features. We handled everything from performance optimization under peak reporting loads to building the audit trail system required for regulatory compliance.
The Challenge
Education compliance software has zero margin for error — schools rely on it for government inspections. Data accuracy, audit trails, and uptime during reporting periods are non-negotiable.
Our Approach
Embedded with the product team as their engineering department. We built incrementally, validating each feature with real schools before rolling out nationally. Shared infrastructure with ClassCharts kept costs down while maintaining separation of concerns.
Tech Stack
FinTech · 3 years
Metis
Financial News Intelligence at Scale
Metis is a financial intelligence platform that ingests news from hundreds of sources worldwide, classifies each article using hierarchical AI models, and surfaces the most relevant signals to investment professionals.
We designed and built the entire pipeline: a distributed crawler that ingests 100K+ articles per day, a multi-stage AI classification system using GPT models for entity extraction, sentiment analysis, and topic categorization, and a search interface powered by Meilisearch for sub-second full-text queries across millions of documents.
The system runs on an event-driven architecture with BullMQ for job orchestration, Clickhouse for analytics queries across billions of rows, and a React dashboard that lets analysts drill down from macro trends to individual articles in seconds.
The Challenge
Financial news moves fast and comes from everywhere. The system needed to process massive volumes with low latency, classify accurately across dozens of categories, and make everything searchable in near real-time.
Our Approach
Started with ingestion reliability — no dropped articles, ever. Then layered AI classification incrementally: basic categorization first, then sentiment, then entity extraction. Each layer was validated against human analysts before adding the next.
Tech Stack
Energy / Technology · 4+ years
Enterprise Energy Platforms
Philips, RWE & Innogy Digital Products
Over several years, we delivered a series of digital products for major European enterprises. For Philips, we built the Metronomis LED Configurator — an interactive tool that lets municipalities and lighting designers configure and visualize LED street lighting solutions before purchase.
For RWE and Innogy (now E.ON), we built MyENMS, an energy management platform that gives facility managers real-time visibility into energy consumption across buildings, with alerting, benchmarking, and reporting capabilities. We also delivered the Innogy Room Planner, a spatial planning tool for office energy optimization.
Each project required working within enterprise constraints: strict brand guidelines, accessibility requirements, multi-language support, and integration with existing backend systems we didn't control.
The Challenge
Enterprise clients need pixel-perfect brand compliance, accessibility standards, multi-language support, and integration with legacy systems — all while keeping the end-user experience simple and fast.
Our Approach
Rapid prototyping to align on UX before building. We delivered interactive prototypes within the first two weeks, iterated with stakeholders, then built production-grade code that met enterprise quality gates.
Tech Stack
Your project could be next
Tell us what you're building. We'll tell you honestly whether we're the right fit — and if we are, we'll start with a free audit so you can see the quality before committing.
Free. No commitment. Takes 30 seconds.