Blackriver is a small, senior engineering company that builds the systems other agencies quote as "not feasible": autonomous AI operations, real-time financial platforms, AI-assisted newsrooms, and the self-hosted infrastructure underneath them. We write the code, we run the servers, and we own the outcome.
A founder answers every enquiry personally. No funnel, no gatekeeping, no discovery-call theatre — we listen to everyone.
Every line below is something we run in production today — for our own products or for clients. If you are a CTO scanning for substance: the systems are named, the stack is named, and you can click through to the live software further down this page.
Autonomous agents that do real work: read a mailbox, query an API, decide, act, and escalate to a human when confidence drops. We build the orchestration, the tool layer, the memory, and — crucially — the evaluation harness that tells you whether it is still working next month.
The entire commercial loop, unattended: an inbound message in a chatroom is qualified, answered, negotiated, closed, invoiced and reconciled — with the payment captured and the receipt filed. Humans supervise the exceptions instead of typing the routine.
The journalist stays in charge; the tooling takes the repetitive work off the desk. Source monitoring, first-pass translation, transcription, image handling, SEO metadata and scheduled publication all run automatically, and every piece is written, judged and signed off by a human. In practice that removes up to 80% of the time spent on repetitive tasks at no cost to journalistic quality.
Machine-payable endpoints built on the HTTP 402 pattern: an agent calls your API, receives a payment challenge, settles it programmatically and gets the resource — no API keys, no manual onboarding, no invoicing round-trip. We build both sides: the paywalled service and the agent that pays.
Data-heavy, real-time web products where the page has to be fast, indexable and correct at the same time: server-side rendering, incremental regeneration, streaming updates, and visualisation that stays readable at a thousand data points. Three live examples are embedded below — judge them yourself.
Real plugin engineering, not page-builder assembly: custom post types, editorial workflows, REST endpoints, caching layers and automation hooks written against the WordPress core APIs — so the site stays fast and upgradeable instead of collapsing under twenty third-party plugins.
Money movement done properly: Stripe integration, subscriptions, metered and usage-based billing, dunning, tax handling, webhook idempotency and reconciliation against the ledger. We treat payment code the way it deserves — as the part that must never silently lose a cent.
On-chain engineering with an execution mindset: smart contracts, DEX integration and liquidity routing, wallet and custody flows, event indexing and on-chain data pipelines. Our own arbitrage system trades against these venues in production, which is a harder test than any audit checklist.
For the systems that have to survive a decade of audits: Spring Boot services, domain-driven design, transactional integrity, batch processing and integration with the legacy estate nobody wants to touch. Boring, documented, testable — which in enterprise is the highest possible compliment.
Everything runs on Linux we administer ourselves — Kubernetes clusters, CI/CD pipelines, monitoring, backups, DNS, and our own mail servers with the deliverability work that actually makes mail arrive. No mystery layer between your product and the metal, and no surprise cloud bill.
We do not buy reach — we operate it. Our own portfolio of publications gives us editorial distribution, and we automate the commercial side of it too: guest-post enquiry handling, placement, publication and billing run as one unattended pipeline.
Purpose-built assistants that live where your team already works — chat, ticketing, mailbox or CLI — wired into your real systems with scoped permissions and a full audit trail. Built on whichever frontier model fits the task, and swappable when a better one ships.
Each panel below is the real, running product loaded directly from its own domain. Scroll it, look at the typography, check how it behaves. This is the fastest honest way to evaluate an engineering partner.
A free market terminal: a global market map, a live financial newswire, symbol research with peer-relative valuation, and a fully transparent model portfolio. It is the hardest category of front-end work — high-frequency numeric data that has to stay legible, accurate and fast at the same time.
A transfer-market news platform built on a continuous data-collection layer: rumours, confirmed moves, fees and club activity are aggregated, deduplicated across sources, scored for reliability and rendered into a browsable, search-indexed archive.
A live fan-voting platform ranking the Greatest Of All Time across multiple sports. Hundreds of thousands of votes have to aggregate into a ranking that updates instantly and cannot be trivially gamed — a write-heavy realtime problem dressed up as a simple leaderboard.
A German-language news portal covering Hungary and Eastern Europe, published continuously since 1854. We run it on WordPress with custom plugins we wrote ourselves. A human journalist stays in full editorial control of every article — what the tooling takes over is the work around the journalism: monitoring sources, first-pass translation, transcription, formatting and metadata. That removes up to 80% of the time spent on repetitive tasks, and none of the editorial quality.
A German travel publication focused on sustainable travel, built on WordPress with custom plugin work and an SEO structure designed for long-tail organic discovery. Part of the publishing network we own and operate — and therefore part of the distribution we can offer clients.
UI components that wrap complex LLM prompts so non-experts can use them. TypeScript, Next.js, Python (Ariadne), GraphQL, Firestore, Google Cloud.
anysolve.ai ↗A search utility that scans public archives and caches to surface open-access versions of articles and academic papers.
nopaywall.net ↗The projects above started as a conversation about something that "probably is not possible". Bring us that conversation.
Talk to us ↗Our most sophisticated in-house system, and the clearest statement of what this team can build. FAAST detects arbitrage sequences across many markets at once, in a search space with far more than two dimensions, and executes them under real latency and real risk constraints. Fully proprietary, wholly owned by Blackriver.
Naive arbitrage compares two prices. FAAST searches for profitable sequences across an n-dimensional space of instruments and venues, using a detection algorithm engineered to be time-optimal — because in this domain, an opportunity found late is not an opportunity.
Price difference alone is a poor signal. We work in the state of the art of financial mathematics: Bregman divergences and Bayesian methods in place of naive geometric distance, statistical arbitrage, and machine-learned signal generation that adapts as regimes shift.
Finding an edge is the easy half. Capital is allocated with fractional Kelly sizing and explicit risk constraints, so the system compounds through drawdowns instead of optimising itself into a single catastrophic bet.
The architecture subscribes to N markets with M pairs concurrently and keeps the whole surface in view at once. Its public output is published at arbitrage-view.com while the system is running — it is not trading at the moment, so treat that page as a window, not a live feed.
We are not selling FAAST. We mention it because it is the honest answer to "can you handle something genuinely hard?" — and because the same people who built it are the people who will build your system.
We are model-agnostic and infrastructure-independent by deliberate policy. If a better model or a cheaper venue appears next quarter, we swap it in — your system should never be hostage to one vendor's roadmap.
A real conversation with an engineer, not a qualification script. You describe the problem; we tell you honestly whether we are the right people.
A narrow first slice that proves the risky part works. You see running software early, not a Gantt chart.
Agile in the plain sense: short cycles, a working increment each time, direct access to the people writing the code.
Deployment, monitoring, backups and on-call on infrastructure we operate — or a clean handover to your team. Your choice.
Blackriver is deliberately small and deliberately senior. The people you speak with in the first call are the people who will design and write your system.
B.Sc. Software Engineering, FH Dornbirn. Leads AI automation, publishing platforms and the FAAST trading system.
M.Sc. Software Engineering, Barcelona Technical University. Leads platform architecture and enterprise engineering.
M.Sc., TU Vienna. Backend engineer, and our specialist in recommendation systems.
We listen to everyone — the fully specified enterprise brief and the half-formed idea that starts with "this is probably impossible, but…". Tell us what you are trying to digitalize. If we are the wrong people for it, we will say so and point you somewhere better.