Systems we built ourselves and are responsible for in production. Client projects covered by confidentiality we describe in conversation.
Own product, built and operated by us
Kontor24 – auction platform
Online auctions in Austria with verified bidders, automatic bidding, audit-ready records and a back office for catalogue, viewing appointments and collection. Bilingual, built entirely in-house and run on our own infrastructure.
Task
An auction platform has to stay correct under load: last-second bids, simultaneous maximum bids, lots closing at the same moment.
Every winning bid creates legally relevant documents: invoice with VAT, collection receipt, the complete history of bids and messages.
Bidders must be identified and able to pay before their first bid; consumer law and GDPR apply from day one.
Implementation
Bidding engine with global resolution of maximum bids; ties go to the earlier bid, and no competing bid is undercut when several bidders are active.
Closing a lot is an atomic, locked transaction; awarding the lot and issuing the invoice are idempotent: a repeated run never produces a second invoice.
Identity and bank verification before the first bid, two-factor login, rate limiting, encrypted storage of sensitive fields.
Tamper-evident event log for bids, messages, invoices and back-office actions.
PDF invoices with VAT logic, collection codes, delivery option, watchlists and email notifications.
Back office for administrators and moderators: catalogue, condition reports, viewing appointments, collection slots, content pages, roles and permissions.
Monitoring with metrics, error tracking and dashboards; backups and recovery documented.
Technologies
Elixir and Phoenix LiveView, PostgreSQL, S3-compatible object storage, Docker on our own infrastructure in Austria.
Own product · licence, rental or adaptation on request
Market data platform and trading automation
Internal analytics platform for crypto and prediction markets (Solana, Polymarket): raw blockchain and order-book data is ingested in real time, aggregated and analysed mathematically; automated trading strategies with backtesting and risk limits run on top of it. Rewritten in Rust with zero-copy processing, because every microsecond counts.
Task
Data streams that never pause: every new block and every order-book change has to be processed, at a volume where conventional databases and row-by-row processing give up.
Analysis has to happen in real time. A signal that arrives late is worthless, and latency creeps in at every link of the chain: parsing, storing, computing, sending.
Strategies must be tested against history before they go live, with the same code that later runs live, otherwise the backtest proves nothing.
Implementation
Ingestion pipeline in Rust with zero-copy parsing of raw data: data is read without being copied and written straight into column-oriented structures.
Aggregation at several time resolutions, metrics and statistical models: volatility, correlations, order-book imbalance, liquidity, forecasts with documented assumptions.
Backtesting engine with identical code paths for simulation and live operation, including fees, slippage and latency.
Execution module with risk limits per strategy, a kill switch, a log of every decision and alerts on deviations.
Measured down to the microsecond: profiling of hot paths, allocation-free processing, no data loss under load peaks.
Operated on our own infrastructure with monitoring, metrics, backups and a documented restart procedure.
Technologies
Rust for ingestion, aggregation and execution; zero-copy parsing; column-oriented storage; Python for research and modelling; operated on our own infrastructure.