Available for staff IC & contract engagements

Senior Staff Engineer

Hands-on engineer and architect, fifteen years of Java and distributed systems in banks and trading firms, currently at JPMorgan Payments, where I lead a team of six building and operating Checkout, the bank's merchant payments product. Now turbocharged with AI, I ship production-grade software fast and still own the whole system: architecture, trade-offs, limits, and what every agent is doing inside it.

Anyone can vibe-code an app. Few can own the codebase behind it at bank scale. On Checkout: POC to production in a year · 12× throughput, 10 to 120 TPS · 12+ months without a high-severity incident · agentic deployment, testing and triage workflows inside change control.

Book a 20-min call Download CV
Juraj Kucera pointing at a “hire me” sign on a rocket at Web Summit
AI-native
Agentic workflows, agent APIs, LLM tool-calling
15 yrs
Java & distributed systems
JPMorgan
Lead a team of 6 on Checkout, 24/7 production
AWS
Certified Developer & Solutions Architect

What I do

Production systems built with agents, and built for agents.

Agents are in my daily engineering loop: Claude Code, Codex and Copilot for building, subagents for research and verification, MCP servers for browsers, cloud and design tools. At JPMorgan Payments I built agentic workflows that automate and harden deployment and testing inside change control, and an agentic triage workflow over Datadog and AWS Athena that shortens incident root-cause analysis.

I also build with LLMs and for agents. For SBM Fizikus, a building-management platform running real apartment blocks, I built an LLM operations assistant that answers questions about live telemetry through tool calls into Postgres. SwingLab exposes market judgment to AI trading agents through an agent API with personal tokens, a Python SDK and a CLI.

Underneath is fifteen years of distributed systems. On Checkout that meant taking the product from POC to production in a year, a 12× throughput increase to 120 TPS, and a staged, controls-first release pipeline that has kept it free of high-severity incidents for over a year. The discipline is the same whether the payload is money or an agent's decision: idempotency, audit trails, deterministic fallbacks, and verification before anything is called done.

  • Agentic engineering workflowsClaude Code, Codex and Copilot; subagent orchestration and plan-verify loops; at JPMorgan, agentic deployment and testing inside change control and Datadog + Athena triage.
  • Agent-native productsSwingLab's agent API, personal access tokens, Python SDK and CLI; versioned scoring that agents consume and humans audit.
  • LLM tool-calling, deployedFastAPI service on the platform's central server, model-agnostic through OpenRouter (Claude, Gemini, GPT), read-only tool calls into Postgres telemetry, per-session cost tracking.
  • Reliability underneathJava, Kafka and Postgres at bank scale; end-to-end release ownership, 12× throughput, 12+ months without a P2, Datadog dashboards, runbooks and RCAs.
Claude Code Codex MCP OpenAI API / OpenRouter Python Java 17–25 Spring Boot Kafka PostgreSQL gRPC React / TypeScript AWS Kubernetes

Side projects

Systems I've designed, built and shipped

Production software outside the day job: a building-management platform with an LLM assistant on top, an agent-native data product, and a scraper people actually use.

SBM Fizikus — wall-mounted climate panel and resident mobile app showing per-room temperature control. ● Live

SBM Fizikus

Smart building management for apartment blocks. Controllers speak MQTT and Modbus to the heating and cooling plant, telemetry syncs to the cloud every 15 seconds, residents get per-room control from an app, and an LLM operations assistant answers questions about live building data through tool calls.

sbmfizikus.com →Python · MQTT/Modbus · AWS · LLM tool-calling
 SwingLab dashboard — market regime tiles for SPY, QQQ, IWM and BTC above a scrollable securities table. ● Live

SwingLab

The judgment layer for AI-driven trading: glass-box signals, scores and market context that agents consume and humans can audit. Regime detection across the major indices, technical state for 9,500+ securities, an agent API with personal tokens, and a Python SDK and CLI.

app.swinglab.site →Java 25 · Spring Boot 4 · PostgreSQL · Python SDK
 Fly Cheap — return flight deals from Central European airports with prices and tracked-price history. ● Live

Fly Cheap

Every Ryanair and Wizz Air return fare out of Vienna, Bratislava, Budapest, Prague, Warsaw, Bucharest and London, refreshed through the day, with the price history behind each one.

flycheap.today →Python · SQLite · Next.js

Writing

Writing on distributed systems

Conceptual essays on the failure modes that decide whether a system stays correct when machines crash and networks blink. Written for engineers who have fought these at scale.

● Published

Two Writes, One Truth

The dual-write problem, why Kafka's exactly-once won't save you, and how the transactional outbox keeps your database and Kafka in sync when money is on the line.

Read on Medium →9 min read · May 2026
 Profile

More on Medium

Distributed systems, payments correctness, and the engineering judgment behind both.

View profile →

Track record

Fifteen years across banks, fintech & low-latency platforms

A selection. Full history and technologies are in the CV.

2023 — Now
JP Morgan Payments · Senior Staff Engineer

Lead a team of six building and operating Checkout, the merchant-facing online payments product (payment form and payment links), with 24/7 production traffic and delivery spread from India to the US West Coast. Took it from POC to production in the first year; own the release lifecycle end to end, with a staged, controls-first pipeline that has kept it free of high-severity incidents for 12+ months; drove a 12× throughput increase (10 to 120 TPS). Built agentic workflows for deployment, testing and Datadog/Athena incident triage.

Java 17SpringKafkaPostgresReactDatadogClaude Code
2023 · 3 mo
Goldman Sachs · Software Engineer, FX TradingContract

Short engagement: built an account rebalancer that auto-adjusts positions against benchmark models.

JavaSlang
2022 — 2023 · 3 mo
Bank of America · API Developer (lead)Contract

Short engagement: lead developer for the API layer of the Metrics Application; migrated Tableau reporting to a Spring Boot + Angular web app.

Java 17Spring BootOpenShiftOracle
2022 · 9 mo
William Hill · Low-Latency Java DeveloperContract

Productionised a greenfield POC in months; hit p95 targets on critical Elasticsearch queries; non-blocking Kafka processing with Apache Flink.

Netty / ReactorKafkaFlinkKubernetes
2020 — 2022
T. Rowe Price · Senior Software EngineerContract

Built an MVP analytics platform for quants and analysts; high-performance gRPC streaming, caching and read replicas for performance, fully remote.

Java 11gRPCAWSAngular
2012 — 2020
Earlier roles · Java developer to senior Java developer

GVC, Kainos for the UK Ministry of Justice, Deutsche Bank and Argus Software in London; BetVictor and AdNovum in Budapest.

Let's talk

If you're building with agents on systems that have to be correct, we should talk.

Open to staff IC roles and select contract engagements in agentic infrastructure, developer tools, AI-native products, and the distributed systems underneath them. The usual way in is a scoped two-to-four-week first engagement with a defined deliverable, B2B through my UK company, remote or in London. Twenty minutes is enough to know if there's a fit.

Book a 20-min call Download CV (PDF)
juraj@jksoftware.co.uk LinkedIn