A technology stack that reflected years of rapid growth and acquisition, in an environment where system changes carry real compliance risk.

Repligen scaled rapidly through organic growth and acquisition, and their technology stack reflected that history: disparate systems, manual integration points, and data workflows that had grown complex with the business.

The challenge was not a single migration or product build. It was a sustained programme of technical modernisation in a highly regulated, operationally sensitive environment where system changes carry validation risk and require careful governance.

Five years of embedded engineering across concurrent workstreams.

Document Intelligence: automated extraction of structured data from life sciences documentation at operational scale, using Azure Document Intelligence. Replacing manual document processing workflows that created operational bottlenecks.

Enterprise platform integration: connecting Repligen's core systems: SAP via OData for operational data, Snowflake as the analytical data warehouse, ServiceNow for IT service management and workflow orchestration, and Microsoft 365 and Teams for the collaboration layer.

Cloud migration: progressive migration of legacy applications and data to cloud infrastructure, with operational continuity and compliance posture maintained throughout.

Two principles that define how this programme runs.

Forward-deployed engineering model. s-team engineers work inside Repligen's teams rather than as an external vendor. This gives the operational and regulatory context to make sound architectural decisions in a GxP-adjacent environment, giving Repligen continuity of technical knowledge across a multi-year programme rather than repeated re-onboarding of new partners.

Progressive modernisation. In a regulated life sciences environment, large-scale system changes carry validation and compliance risk. Each workstream is scoped to maintain operational continuity, with changes validated before the next phase begins.

An ongoing partnership, entering its sixth year.

Forward-deployed engineers with deep domain and operational context: this is how s-team approaches complex, long-horizon technical programmes where continuity of knowledge is as valuable as technical capability.

Three languages. One knowledge layer. No engineering involvement to maintain it.

This operator's contact centre agents handle thousands of customer queries daily across three languages: English, Arabic, and Kurdish (Sorani). Their existing knowledge base was static documentation. Agents searched manually, results were inconsistent, and Arabic search quality was poor.

Most enterprise knowledge tools treat Arabic as a secondary language, producing mismatched results, particularly with dialect variation and transliteration. Before committing to a full platform build, the operator needed to answer one concrete question: could AI-powered knowledge retrieval work reliably across all three languages, in a production-grade architecture, with governance controls their content team could operate without engineering support?

A working system, not a mock-up.

We designed and delivered a fully functional proof of concept, not a mock-up built to impress in a demo room. The architecture was production-ready from day one.

The system covers the full intelligence cycle: natural language query understanding in English, Arabic, and Sorani Kurdish; ranked article results with copy-ready agent responses; automatic knowledge gap detection and logging; and live re-indexing triggered by CMS publish events with no engineering involvement. Four distinct access tiers (internal staff, contact centre agents, customer-facing, and API/LLM consumers) served from one knowledge layer with no duplication.

Three decisions that shaped the architecture.

Multilingual embeddings as first-class citizens. Rather than translating Arabic content into English for embedding, each article is embedded per locale. Arabic queries retrieve Arabic results: semantically matched, not string-matched. Gemini Embeddings provided multilingual capability across Arabic, English, and Kurdish without custom model training.

Live re-indexing via webhook. When a content editor publishes in Strapi, the search server receives a webhook, re-embeds the content, and updates the index in real time. No restart, no delay, no engineering ticket. Maintainable by content teams, not just engineers.

Knowledge gap detection as a core feature. Queries returning low-confidence results are automatically flagged and logged in a prioritised gap report. Agents see a gap indicator; the content team sees an actionable list. Closing a gap is a one-step workflow: write the article, publish, and the system validates the improvement immediately.

End-to-end multilingual AI retrieval, in a working system.

The POC demonstrated accurate Arabic and Kurdish query understanding, ranked results under 200ms, live re-indexing, and a knowledge governance workflow the content team could operate without touching code. The architecture is designed for full production deployment with no structural changes required, giving the operator confidence to commit to a full platform build on proven foundations.

Enterprise software serves large carriers. Regional operators manage complex operations with tools that were never built for them.

Enterprise aviation software is designed for large carriers and priced accordingly. Regional operators (charter companies, small airlines, and specialist aviation businesses) manage complex, regulation-heavy operations with tools that were never built for them. Most cobble together spreadsheets, email, and disconnected systems to run businesses where precision and compliance are non-negotiable.

AirTrail's mission: purpose-built operational software for regional air operators. Accessible enough for small teams, capable enough for serious operations.

Five modules. Replacing paper, email, and disconnected systems with a single operational record.

Flight and maintenance logging: actual flight times, recency information, trend tracking, and defect recording. Replacing the paper-based workflows most regional operators still rely on, with a complete and auditable digital record.

Crew and AME sign-offs: digital workflow for crew sign-off and Aircraft Maintenance Engineer approvals. Every entry creates an immutable record with timestamp and signatory. No paper, no ambiguity, no gap in the audit trail.

Regulatory compliance management: certification tracking, documentation workflows, and audit readiness for aviation regulatory requirements. Built to reduce the compliance burden, not add to it.

Charter booking and customer management: booking workflows for charter and private air travel, with customer records and communication tooling integrated into the operational picture.

Operational dashboards: leadership visibility across fleet status, schedule, and maintenance in real time. One screen, not five systems.

An active partnership. Still building.

AirTrail continues to evolve the platform with s-team as the development partner, expanding capabilities and refining the product based on operator feedback in the field.

A fragmented ecosystem with no shared trust layer.

Tourism platforms are fragmented: travellers, venues, creators, local businesses, and transportation hubs operate in separate digital ecosystems with no shared trust layer. Kbean's founding vision was to build the connective infrastructure: a platform where a traveller's identity, content rights, location context, and community connections operate on a single authenticated layer.

The platform needed to span mobile applications, blockchain rights management, media streaming, marketplace mechanics, and eventually live deployment in international airport infrastructure.

Four distinct products over seven years, all on one evolving architecture.

Kbean Marketplace: a rights-managed digital art and media marketplace powered by patented MicroNFT® technology, enabling micro-priced fractional digital rights with transparent royalty mechanics for creators.

Inssa: location and time-locked digital connections delivering content, offers, and gifts to travellers at the right moment and place. Now live at Denver International Airport, connecting travellers with local businesses and cultural institutions.

Challenge: gamified creative challenges that protect every submission on-chain and celebrate winners on a shared stage. Deployed with arts institutions including the Minhwa Challenge (Korean traditional art) and the SALT student film festival.

Kbean Black: immersive heritage and cultural venue experiences transforming museums, restaurants, and cultural institutions into verifiable digital experiences.

Three decisions that shaped the architecture.

MicroNFT® architecture. Standard NFT mechanics are built for high-value, single-owner digital assets. Kbean's model required fractional, micro-priced digital rights: a bespoke token model built around micro-licensing rather than ownership transfer, outside standard ERC-721 patterns.

Architectural continuity across seven years. Kbean has pivoted significantly, from digital media marketplace to smart tourism to airport deployment. The architectural decisions made in year one were designed to accommodate evolution. The Inssa deployment at Denver International Airport runs on the extended foundation of the original architecture, without a rebuild.

Multi-platform from the start. React Native for iOS and Android, with streaming integrations covering Fire TV, Roku, and Apple TV during the media platform era. Built to serve both consumer and venue-operator use cases from the same codebase.

Live at Denver International Airport.

Inssa connects travellers with local businesses and cultural institutions at one of the ten busiest airports in the United States, connecting travellers with local businesses and cultural institutions at scale. The platform has been recognised through partnerships with arts institutions across the US and continues in active development with s-team as the architecture partner.

The gap between AI's analytical capability and its operational usability in regulated environments is a transparency problem.

AI adoption in regulated industries stalls at the same point every time: the model produces a score and nobody can explain why. A compliance officer cannot act on an AML alert they cannot justify to a regulator. A risk manager cannot defend a people-risk decision in a legal context.

The gap between AI's analytical capability and its operational usability in regulated environments is a transparency problem. That was the founding insight behind Talisai. Explainability needed to be a core architectural principle, not an audit log bolted on afterwards.

Talisai's full SaaS platform, built from scratch, across three product modules.

AML for Crypto Exchanges: real-time transaction monitoring with risk scoring and full rationale. Compliance officers see not just the alert but the specific risk indicators that triggered it, each weighted and traceable. Configurable risk indicator libraries that evolve with regulatory requirements without a rebuild.

People At Potential Risk (PAPR): out-of-the-box risk factor monitoring for customers, business partners, and employees. Hybrid deterministic and ML models: the explainability of rules combined with the pattern-recognition of learned models.

Supply Chain Vendor Risk: vendor security and privacy compliance monitoring against ISO 27x, SOC2, and HIPAA, with anomaly detection and evidence chains for audit.

Three architectural decisions that made explainability possible.

Hybrid model architecture. Neither pure rules nor pure ML was sufficient. Rules are explainable but rigid; ML learns patterns but is opaque. The platform combines both: deterministic scoring for the explainable baseline, ML for patterns the rules don't catch. Each component of the final score is traceable to its source.

Evidence chains as a product feature. Every risk score generates a traceable evidence record (specific inputs, model weights, and rule triggers) queryable for regulatory response without manual investigation. Designed into the data model from day one.

SaaS or on-premises deployment. For clients with data sovereignty requirements, the platform deploys on-premises with no data leaving the customer environment. The deployment model was a product decision, not an afterthought.

A production platform recognised in the market from launch.

Talisai went to market with a production SaaS platform recognised by Enterprise Security Magazine as a Top 10 Provider of Automation Security for 2020. The platform serves financial services, healthcare, and enterprise risk management clients requiring accountable, auditable AI.

The execution has to match the brand.

Luxury brands entering digital collectibles face a specific problem: the execution must match the brand. A standard NFT drop built on an off-the-shelf platform doesn't belong next to a $50,000 watch.

Hublot's "first, different, unique" philosophy needed to carry through into the digital experience, without compromise. That's not a creative brief. It's a technical constraint.

Web3 and NFT infrastructure engineered for luxury.

Web3 and NFT infrastructure for Hublot: a digital collectibles experience engineered to sit alongside the physical product, not beneath it. Smart contracts on Ethereum ERC-1155, a TypeScript React frontend, Node and NestJS backend, Metamask integration, and Azure App Services infrastructure: the full stack, built to luxury standards.

Architecture chosen for fit, not speed.

Infrastructure chosen for fit, not speed. ERC-1155 supports multi-token contracts: multiple asset types under one contract address, with provable scarcity and full on-chain ownership history for every collectible. The technical architecture was chosen because it was right for luxury collectibles mechanics, not because it was fast to implement.

Hublot positioned as a serious participant in the Web3 luxury space.

Hublot's digital collectibles experience launched and positioned the brand as a serious participant in the Web3 luxury space, a market where most participants are still figuring out what "luxury" means on-chain.

Variable demand. Fixed supply. No infrastructure for the market in between.

As hybrid work grew, businesses faced a recurring problem: permanent office space didn't match variable workforce patterns. On peak days, overcrowded. On quiet days, empty. When project teams needed temporary space (a client meeting, a sprint week, an offsite) the options were expensive serviced offices or unsuitable co-working environments with no granular booking control.

Looffiz set out to build the infrastructure for a flexible office market: a two-sided platform connecting businesses with surplus workspace to those needing it.

A two-sided marketplace built for both supply and demand from the first line of code.

The full Looffiz platform: a two-sided marketplace designed around granular, flexible booking. Space providers list available offices with capacity, amenities, location, and pricing. Businesses browse, filter, and book in real time, with billing handled end-to-end in the platform.

Space listing and management for providers: floor plans, amenity tags, availability calendars, pricing tiers. Granular booking engine supporting hourly, daily, weekly, and monthly reservations on one shared availability model. Real-time booking confirmation. Integrated payment processing. Two-sided review system. Admin tooling for marketplace operations and dispute management.

One architectural decision that prevented the most common marketplace failure.

Two-sided marketplace mechanics from the start. Building for both supply and demand simultaneously is a product and engineering challenge. Provider-side tooling and buyer-side tooling were built on shared data models, ensuring availability and pricing was always consistent across both sides of the market. A common failure mode in marketplace builds (divergent data models) was designed out from day one.

Launched complete. Owned outright.

Looffiz launched in the Netherlands fully operational: listings, availability, payments, reviews, and admin all live at launch. The client owns it outright. No lock-in, no ongoing dependency on s-team.