Selected work

Real systems. Real impact.

A selection of AI systems we’ve designed and shipped with our partners — across agents, retrieval, voice, applied ML, and platform engineering. Sensitive details are abstracted; the business and technical story is real.

Start a project

Agents

An agent that turns a business scope into a deployed service

A production R&D system that takes a business scope and produces a deployed backend — generating agent graphs, tool configs, and an integration-ready API surface.

Agent orchestrationTool callingStructured planning+1

View case study

Agents2026

Business scope to a deployed FastAPI service, autonomously

Agents

Deep-research agents for decision-ready reports

Agents that retrieve, read, and synthesize information into structured analyses — with predictable structure, grounded outputs, and repeatable quality.

Agent orchestrationEnterprise searchDocument intelligence+1

View case study

Agents2026

Decision-ready reports with a predictable structure

Platform

An evaluation & regression suite for LLM features

An internal framework that benchmarks agent outputs against gold standards, tracks regressions across prompt, model, and logic changes, and makes quality trends visible.

Evaluation datasetsAutomated scoringRegression tracking+1

View case study

Platform2025

Regressions caught before they reach users

Retrieval

A multi-tenant documentation & troubleshooting assistant

An enterprise assistant with retrieval-augmented answers, built for multi-tenant usage with scalable retrieval and persona / access patterns.

RAGMulti-tenant retrievalPersona & access control+1

View case study

Retrieval2025

Grounded answers from large, tenant-specific knowledge

Retrieval

Hybrid semantic retrieval where precision and recall both matter

Semantic and hybrid search combining keyword precision with semantic recall — designed to be tuned and observed, so retrieval quality is something you can measure.

Hybrid searchVector retrievalRe-ranking+1

View case study

Retrieval2025

Keyword precision and semantic recall, balanced

Voice & Multimodal

Real-time voice agents across 40+ languages

Natural, human-like voice agents for sales and support — low-latency, multilingual, and integrated with enterprise data, including Gemini Enterprise on Google Cloud.

Real-time voiceLow-latency pipelinesMultilingual+1

View case study

Voice & Multimodal2026

Natural, human-like interaction in 40+ languages

Applied ML

Forecasting and risk models that feed operations

Time-series forecasting for retail checkout volumes and sales trends, plus credit scoring and risk models for banks and large e-commerce platforms — wired into real planning systems.

Time-series forecastingCredit scoring & riskFeature pipelines+1

View case study

Applied ML2024

Forecasts that feed operational planning directly

Computer Vision

Real-time in-cabin monitoring on the edge

Computer vision for in-cabin monitoring — pose, emotion, and object detection from rear-view-mirror cameras, optimized to run in constrained embedded environments.

Computer visionEdge / embedded AIReal-time inference+1

View case study

Computer Vision2024

Real-time perception on constrained hardware

Platform

ML infrastructure for a mobility platform at scale

End-to-end machine-learning infrastructure and lifecycle management for one of Europe’s largest mobility and ride-hailing platforms — massive-scale ingestion and deployment across sectors.

MLOpsData ingestion at scaleModel deployment+1

View case study

Platform2024

ML lifecycle managed end to end

Let’s talk

Your problem probably looks like one of these.

If you’re shipping AI into a real product and need it to hold up, let’s talk about how we’d build it.

Book a Discovery Call See our work