Available · Accepting projects now
AI · Data · Cloud
We build data and AI systems end to end: from pipelines and cloud architecture to the final interface.
The engine
We built and run our own analytics platform: +1,000 monthly users, +100 paying, 50M+ data points processed. It handles live multi-source ingestion, opportunity detection and position tracking, in production, under real load, with paying users who notice immediately when something breaks.
That engine is the base we adapt for client work. Five of our nine clients got a decision panel shaped to their own criteria, their own filters and their own KPIs, on infrastructure already proven in production.
Weeks instead of months, without the risk of a first deployment.
Agents and RAG grounded in your own material, not generic model knowledge. Automatic classification and auditing, anomaly and fraud pattern detection, first-line support that actually knows your business. In production, with real users, not a demo.
Decision panels built on a proven analytics engine: live data from multiple sources, opportunity and exception finders filtered by your own criteria, position tracking and alerts. Adapted to each client rather than rebuilt from zero.
Backtesting engines, algorithmic strategy design, scenario modelling and valuation analysis. In-house formulas, validated against historical data before any capital is at risk. This is where our banking background does the work.
From the first line of code to the paying user: data model, roles and authentication, payments, automated onboarding, analytics. Serverless or dedicated, decided on cost and load.
Version control with full change history · Automated dependency and credential scanning · CI/CD with tests before every release · Architecture, environment variables and deployment procedures documented.
Nine clients delivered. Two are set out in full; the rest in brief.
Before
Days, by hand
After
Minutes
Regulatory reporting automation · Regtech · Compliance
Producing a full tax position report meant reconstructing every transaction by hand: identifying its source and nature, classifying it, and reconciling P&L. A full day's work, sometimes several. ChainTaxFlow does it in minutes and outputs a report ready to export and feed into filing platforms.
Python in a serverless environment means paying only for the seconds each analysis runs. For intermittent loads the cost per report is negligible against a permanently running server. And building our own indexing engine, rather than licensing a generalist one, is what lets the report cover the non-standard cases instead of leaving them unclassified.
The agent supports and speeds up recurring queries. It does not replace professional tax advice.
Before
Hours across tabs
After
Ten minutes, one screen
Financial analytics platform · Hong Kong · Trading & investment analytics
Evaluating a strategy meant jumping between platforms and gathering indicators one at a time, an afternoon of work before putting capital at risk. Quantik processes thousands of historical data points per simulation and returns drawdown, Sharpe ratio and win rate calculated, synthesised and explained on a single screen.
The client needed an all-in-one architecture on a very light server with no third-party dependency: database, authentication and admin panel. Self-hosted PocketBase does exactly that and prototypes fast when the project doesn't need to scale to a mass user base. The tool answers the problem, not the habit.
Usage analytics showed which parameters almost nobody touched. They moved to an advanced panel, leaving the main screen with only what was actually used.
Before
Assembled by hand
After
One panel, own filters
Investment decision dashboard · Asset Management
Investment decisions stopped depending on data assembled by hand. The team filters by its own criteria and sees the opportunities that match, as conditions change.
Before
Found out of range late
After
Alerted on drift
Position tracking & alerting · Asset Management
Out-of-range positions stopped being something discovered late. R2R sees the state of every position and where capital would work harder, in one place.
Before
Judged before tested
After
Backtested first
Algorithmic strategy design · Quantitative engineering
A production-ready strategy with backtested performance behind it, letting Gliquid optimise capital efficiency before committing real money.
Before
Argued with narrative
After
Modelled across scenarios
Quantitative valuation modelling · Scenario analysis
Gave Plutus a quantitative basis to evaluate sustainability and to argue value to investors with numbers rather than narrative.
Before
Theory, no live data
After
Live data in the lesson
Custom analytics dashboard for education · plus two further training platforms
A personalised analytics panel for each academy, carrying the KPIs and metrics of its own programme. Real-time data, opportunity finders, and data pipelines integrating live analytics.
Students work with professional-grade market data inside the platform where they learn, instead of theory detached from live conditions. Three training companies now teach with a panel that matches how they actually operate.
Weeks 1-2
Workshops with your team. We map systems, data, bottlenecks, objectives. You get a roadmap prioritised by impact.
Weeks 2-5
Architecture, data model, API contracts, navigable prototype. Validated before production code.
Weeks 5-18
Two-week sprints, live demos, CI/CD from day one, tests and documentation included.
Ongoing
We operate it, transfer the knowledge, leave your team owning a platform they understand.
Team
Every engagement is led by someone who has broken production and fixed it. Behind the three of us, a network of specialists we bring in per project.
Head of Engineering · Cloud, Data & AI
Statistician and data engineer. Designed end-to-end cloud architecture and ETL pipelines. Banco Sabadell (risk) and Albeit Capital (BI & strategy).
Head of Data Science · ML, Risk & Quantitative Modelling
Applied statistician. Independent validation of risk models at Bankinter (IFRS 9, ICAAP, ILAAP). Best Master's Thesis Award, UCM. Turns complex data into decisions.
Head of Product Engineering · Frontend & UX
Frontend and UI/UX. Custom dashboards, pixel-perfect layouts and scalable web applications.
Two statisticians from Universidad Complutense and a full-stack developer. Guillermo spent three years validating risk models at Bankinter under IFRS 9, ICAAP and ILAAP. Roberto works on credit risk and scoring in Banco Sabadell's data hub, after two years at Albeit Capital automating reporting and building the dashboards behind investment decisions. Regulated environments, where a model has to be defensible and not just accurate.
Contact
If your challenge is technical, an engineer answers, not a salesperson.