PAX·CODE — home

EN / PL

Book a call Menu
  1. 00
  2. I
  3. II
  4. III
  5. III.1
  6. IV
  7. 05

SYS / 00 — PAX·CODE

Software architecture that holds.

Software architecture and engineering for backend, data and AI systems — eight years of keeping production in order.

Book a call See the work
8+ years in production
~400B records scanned
3+ years of AI in fintech

Scroll to build

SYS / 01 — Power & cabling

I.Built from the backend up.

Since 2018: freelance work, a year of QA automation, then full-stack engineering in product teams. Today PaxCode contracts B2B — Node.js and TypeScript services, AWS, and data at a scale most teams never reach.

  • F / 01 Backend
  • F / 02 Data
  • F / 03 AWS

SYS / 02 — Devices

II.One drum per system in production.

Each project below carried real load, with decisions and trade-offs documented.

  1. Drum 03 AI pipelines for financial advice 3+ years for a UK AI fintech. Pipelines that turn adviser meetings into finished documentation. Read case study
  2. Drum 02 No-code pipeline builder Proposed and built for the same client. Hundreds of client firms use it to configure their own AI pipelines. Read case study
  3. Drum 01 Data leak scanner ~400 billion records on Athena, S3 and Parquet. Designed and run solo. Read case study
All work

SYS / 03 — Capitulum

III.AI that does the paperwork.

The current focus is AI in production: LLM pipelines over meeting transcripts, prompt design, multi-modal processing and caching that keeps each run fast and affordable. Plus the tools that let non-engineers build pipelines of their own.

  • I. Architecture & review
  • II. Backend & data
  • III. AI / LLM pipelines
  • IV. Product & UI/UX design
  • V. Senior engineer on contract
  • End to end — from the data model to the interface, one owner.
Services

SYS / 03.1 — Capitulum · core

III.From raw input to a finished answer.

One adviser meeting, followed through a production AI pipeline — twenty steps from a recording to a reviewed report in the client’s CRM. Keep scrolling to run it.

Open any step for the business and the engineering view

Overview · 20 steps
01 — Collect → raw input → clean, queued events
  1. 01 Sources audio · pdf · api
  2. 02 Parse STT · OCR · diarise
  3. 03 Normalise schema · metadata
  4. 04 Validate dedup · checks
  5. 05 Event queue SQS · retries · DLQ
02 — Process ← events → searchable knowledge
  1. 06 PII redaction NER · rules · GDPR
  2. 07 Chunking semantic · overlap
  3. 08 Embeddings batch · cache
  4. 09 Hybrid index vector · BM25
  5. 10 Knowledge base versions · filters
03 — Reason → task → grounded draft with citations
  1. 11 Agent orchestrator plan · act · observe
  2. 12 Retrieval (RAG) top-k · rerank
  3. 13 Context engineering tokens · memory
  4. 14 Tools MCP · functions
  5. 15 LLM gateway routing · cache
04 — Deliver ← draft → checked, delivered answer
  1. 16 Guardrails input · output
  2. 17 Structured output schema · retry
  3. 18 Human review human in the loop
  4. 19 Delivery SSE · docs · API
  5. 20 Client systems CRM · webhooks
Observe & evaluate Tracing · OpenTelemetry Evals · golden sets, LLM judge Cost & latency per run Monitoring · drift, alerts Feedback → re-index
Example pattern · fictional data · not a client system End to end: architecture, backend, AI pipelines and the UI/UX on top

SYS / 04 — Pax

IV.Order in complex systems.

Pax Romana meant two centuries of stability. The goal here is more practical: systems that stay understandable as code, data and AI pile up.

  1. 01Decisions written down, trade-offs stated.
  2. 02Numbers over adjectives.
  3. 03The simplest design that survives the next order of magnitude.

Latest writing

All posts

[DO UZUPEŁNIENIA: first posts]

SYS / 05 — Contact

Let’s look at your system.

A 30-minute call: what you’re building, where it hurts, and whether PaxCode is the right fit. Based in Poland, working remotely with teams in the UK and EU.

Book a call Write to hello@paxcode.it