Ideas become valuable when they create measurable change.

The challenge: Manual AP processes — slow, costly, and error-prone. Fraud risk from fake vendors, IBANs, and VAT manipulation. Designed to address GDPR requirements and enterprise security controls, with SOC 2-aligned practices where applicable.

Our solution: Tiered multi-agent pipeline (N1–N4); Doc AI for SIRET, Factur-X, and BTP VAT; 4-layer fraud detection built in; cloud-agnostic and Kubernetes-native architecture; GDPR and SOC2 compliant from day one.

Results & impact: Intelligent automation at scale · Lower fraud exposure on every invoice · French/EU compliant and audit-ready · No cloud lock-in · ~80% lower LLM token cost · Kickoff to production-grade in 2 months.

"Manual AP replaced by a fraud-aware, EU-compliant, multi-agent platform — built in two months."

Architecture: Kubernetes-native · cloud-agnostic

AI: GPT-4.1 · LangChain4j

Data: PostgreSQL · pgvector

Integration: Kafka · MCP

Outcome: ~80% lower LLM token consumption

The challenge: Unused & underutilised ServiceNow licences, duplicate and outdated Configuration Items in CMDB, limited visibility into platform adoption, weak governance around assets and reporting, and manual effort to identify cost optimisation areas.

Our solution: Analysed licence allocation and identified inactive/duplicate licences; assessed CI quality and removed obsolete CMDB records; improved CMDB governance and data quality; reviewed module adoption trends; delivered dashboards with continuous optimisation; established a governance framework for licence management.

Results & impact: Reduced unnecessary licensing costs · Improved Configuration accuracy · Increased platform visibility · Enhanced reporting through clean CMDB data · Improved decision-making through actionable dashboards · Created a sustainable framework for ongoing platform optimisation.

"From unclear platform usage to a governed, cost-optimised ServiceNow environment with full visibility and control."

Platform: ServiceNow · CMDB · ITAM

Optimisation: Licence Management · Usage Analysis

Governance: CMDB Quality · Asset Governance

Reporting: Executive Dashboards · Adoption Analytics

Outcome: Lower licence cost · Improved platform visibility

The challenge: Only a code repository — no documentation. Mission-critical legacy .NET system where changes required scarce specialists. Slow, risky updates with knowledge at risk.

Our solution: Reverse-engineered the whole system; documented functions, architecture, and stack; optimised code and cut technical debt; enabled agentic-AI development workflow.

Results & impact: Black box → fully documented · Faster, lower-risk feature changes · Client updates the system itself · Proven business logic preserved.

"From undocumented black box to an AI-augmented platform the client can evolve on its own."

Architecture: Enterprise .NET · C#

Modernisation: Static analysis · Code refactoring · Agentic-AI dev tooling · Documentation

AI: Agentic AI Development Tooling

Engineering: Reverse Engineering · Technical Documentation

Outcome: Lower technical debt · Faster, lower-risk changes

The challenge: Complex, old and large codebase. Angular 2 — end-of-life and insecure. Outdated Java SDKs & libraries. Inflated, un-optimised AWS bill.

Our solution: Migrated Angular 2 → 18 incrementally; upgraded Java to current LTS; refactored for structure & documentation; right-sized AWS infrastructure and cut waste (FinOps).

Results & impact: 40% lower AWS cloud bill · Faster, safer feature delivery · Reduced security exposure · Future-proofed, maintainable stack.

"A four-year-old, hard-to-change ERP, re-platformed to modern tech — at 40% lower cloud cost."

Frontend: Angular 18 · TypeScript · RxJS

Backend: Java (LTS)

Cloud: AWS

Engineering: CI/CD · Code Refactoring

Outcome: 40% lower AWS cloud spend

The challenge: EV charger installation and servicing involved multiple disconnected activities, including finding qualified professionals, scheduling installations, managing change requests, processing payments, and resolving support issues. The client needed a centralized and scalable platform for residential and commercial customers.

Our solution: Developed a mobile and web platform featuring secure login, dealer search, installation scheduling, provider reviews, image uploads, change-order management, service completion details, surveys, and multiple payment options. Real-time chat and notifications improve customer-provider communication. A RAG-based AI chatbot answers queries, creates support tickets, and transfers urgent issues to a human support agent.

Results & impact: Simplified the complete EV charger installation and service journey through one connected platform. Improved workflow visibility, customer communication, payment processing, and support management. The project achieved 15% cost savings against the estimated budget and was delivered ahead of schedule.

"A connected EV service platform that manages everything from charger discovery and installation to payments and ongoing customer support."

Applications: Flutter · Dart

AI: ChatGPT · RAG · LiteRT

Backend & Data: Python · Express.js · MySQL · Redis

Cloud & Integration: AWS · Stripe · Firebase · Elasticsearch · Twilio

Outcome: 15% below estimated budget · Delivered ahead of schedule

The challenge: Consumers lacked a unified way to monitor electricity consumption, understand their carbon footprint, track changing energy prices, and manage payments. The platform also needed to connect multiple smart-home and energy devices while securely processing real-time data from different systems.

Our solution: Developed a mobile and web-based utility platform featuring real-time electricity and carbon-usage dashboards, location-based pricing notifications, personalised payment limits, transaction history, recurring payments, instant invoices, rewards, gift coupons, feedback, and customer support. The middleware integrates live energy-pricing APIs, smart-grid systems, and IoT devices from Ecobee, Honeywell, Google Nest, and Tesla. It can automatically manage connected devices based on real-time energy prices.

Results & impact: Enabled consumers to make informed energy decisions, automate connected devices, and manage payments through one platform. The solution supports 100% renewable energy adoption, helps reduce carbon footprints by up to 70%, and enables potential electricity bill savings of up to 40%.

"A connected energy ecosystem that turns real-time consumption and pricing data into smarter, greener, and more cost-effective energy decisions."

Applications: Flutter · Angular

Backend: Java · Spring Boot

Data: MySQL · SQL Server · MongoDB

Integration: IoT · Smart Grid · REST APIs

Outcome: Up to 40% potential electricity-bill savings

The challenge: AI professionals face an overwhelming volume of news, research, and technical content published across multiple sources every day. Manually collecting, reviewing, categorising, and summarising this information was time-consuming, costly, and difficult to scale. Users also required content tailored to their interests and level of expertise.

Our solution: Developed a mobile application and backend content management system powered by classical machine learning, GenAI workflows, AI agents, and human-in-the-loop validation. The platform automatically sources, filters, classifies, tags, summarizes, and recommends content. It also generates supporting images and creates personalised feeds based on each user’s interests, selected AI categories, and expertise level.

Results & impact: The platform processes approximately 300–500 articles every day from more than 70 sources while maintaining an average GenAI processing cost of approximately $0.30 per day. Automated workflows significantly reduce manual content-management effort while delivering relevant, accessible, and personalised AI knowledge to users.

"A scalable multi-agent content platform that transforms complex AI news and research into personalised, easy-to-consume insights."

Application: Flutter

AI: Flutter · GPT-4 · GPT Image 1 · CrewAI · Machine Learning

Backend: Node.js · Python

Data & Cloud: PostgreSQL · AWS

Outcome: 300–500 articles/day · ~$0.30/day GenAI processing