Senior Software, Data & AI Engineer

I build systems that make complex data useful, reliable and fast.

Three decades in software engineering, including more than seven years focused on data platforms, streaming and production AI. I turn ambiguous business problems into durable systems with measurable outcomes.

See selected work Start a conversation

Software & Data Platform résumé Data & AI résumé

Based in Germany Worldwide opportunities Remote preferred Hybrid nearby · no relocation

10×faster data processing

34%lower Snowflake cost

72%fewer data incidents

60%lower AI operating cost

Selected work

Systems, not demos.

These are compact, public-safe views of private production work. The examples show the engineering decisions and operating principles without exposing credentials, financial data or proprietary rules.

Flagship public repositoryOpen source

Production Data & AI Decision Platform

A production-oriented reference implementation for turning imperfect operational data into typed, evaluated decisions. Python and Go services, Prefect orchestration, an optional Jev adapter, graph execution, model routing, deterministic release gates, redacted telemetry, Docker, CI and synthetic data.

Fail-closed release gates
Cost-aware model routing
Containerized tests and benchmarks

Sources→Typed classifier→Decision graph→Evals→Release gate

Explore the repository →

AI engineering2026

Reliable AI Work Harness

A graph-based execution platform where agents, models and tools operate behind versioned contracts. Routing, retries, validation, telemetry and human decisions are explicit—so AI workflows can be tested, audited and improved instead of merely observed.

Bounded retries
Redacted traces
Policy-driven routing

Simplified, public-safe pattern

class StepPolicy:
    attempts: int
    timeout: float
    side_effect: "none | idempotent | unsafe"

    def validate(self):
        if self.attempts > 1:
            require(self.timeout)
            reject(self.side_effect == "unsafe")

Read the case study →

Product engineering2025–present

Investment Manager

A private decision and operations platform spanning portfolio reporting, options workflows, tax-supporting outputs, reconciliation and controlled execution. PostgreSQL is the system of record; critical changes use preview, transaction and audit boundaries.

99.8% data coverage
90 → <10 min reconciliation
Auditable operations

Operational boundary

with transaction():
    plan = preview(command)
    validate(plan, current_state)
    result = apply(plan, idempotency_key)
    audit(result, actor, source_version)

Read the case study →

Data platformProduction

Prefect Market Data Platform

An orchestration layer for market data, pricing, analytics and strategy workflows. It combines bounded flow policies, official-source backfills, fallback providers, observability and fail-closed finalization for business-critical outputs.

Multi-source ingestion
Recoverable backfills
Fail-closed gates

Resilient flow pattern

@flow(timeout_seconds=policy.deadline)
def ingest_market_data(date):
    symbols = guarded(fetch_symbols, policy.fetch)
    prices = parallel(fetch_prices, symbols)
    validate_coverage(prices)
    finalize_atomically(date, prices)

Read the case study →

Data engineeringCareer impact

Faster, cheaper, safer data

Across streaming, Snowflake and Spark platforms, the recurring work is the same: make freshness a product capability, make cost observable and make reliability part of the architecture—not a cleanup phase.

4× throughput
34% lower cost
72% fewer incidents

Data product contract

contract = {
  "freshness": "near real-time",
  "idempotent": true,
  "replayable": true,
  "owner": "named",
  "cost_visible": true
}

Read the article →

Engineering notes

Decisions behind the systems.

Short, practical articles about reliability, cost, architecture and operating production systems. Each is written to be useful on its own—and easy to share on LinkedIn.

01 / AI SYSTEMS

Reliable AI needs contracts, not just prompts

Why routing, bounded retries, validation, redacted telemetry and explicit side-effect safety belong in the platform.

Read the article →

02 / ORCHESTRATION

Prefect or Airflow? A workload-based decision

Why faster Python delivery, flexible execution and a smaller infrastructure footprint fit this platform—and when Airflow would be stronger.

Read the article →

03 / AUTOMATION

How I reduced my daily trading routine to 30 minutes

How data ingestion, strategy pipelines, validation and a controlled review surface turned hours of fragmented work into one focused routine.

Read the article →

04 / IMPACT

How data platforms become faster, cheaper and safer

The engineering levers behind 4× throughput, 34% lower warehouse cost and 72% fewer incidents.

Read the article →

05 / DECISION SUPPORT

From data signals to operational opportunities

How Harness, graph workflows, LLMs, evals and the type-safe JEV classifier can support intraday decisions without surrendering control.

Read the article →

06 / STREAMING

From PostgreSQL CDC to near-real-time Snowflake

Why Spark Structured Streaming belonged between Pub/Sub and Snowflake—and which PostgreSQL WAL problem it could not solve.

Read the article →

07 / OBSERVABILITY

When Prometheus outgrew product observability

How Snowplow, Snowflake Streams and Grafana scaled analysis of users, links, pages, downtime and system behavior.

Read the article →

Expertise

Depth across the full system.

I work from product intent through architecture, implementation, data contracts, deployment and operations. The goal is not a fashionable stack; it is a system a team can trust.

Software engineering

Python, Go, TypeScript, Clojure, distributed systems, APIs, event-driven design, CQRS and transactional workflows.

Data platforms

PySpark, Spark Streaming, Snowflake, dbt, Airflow, Prefect, PostgreSQL, CDC, data modeling and observability.

Applied ML & AI

LangGraph, agent systems, routing, evaluation, telemetry, retrieval, model-cost control and safe production integration.

Technical leadership

Architecture, migrations, incident reduction, cost optimization, mentoring, stakeholder alignment and platform strategy.

Earlier analytical work

Useful foundations, kept in context.

Selected studies that show long-standing interest in markets and statistical reasoning. They support the story; they no longer define it.

Stock market anomalies

Exploratory analysis of anomalous price behavior in Brazilian equities.

View on GitHub →

Trend or mean reversion?

A study of asset behavior after positive and negative daily candles.

View on GitHub →

Let’s build something dependable

Senior engineering for complex software, data and AI systems.

Open to Senior and Staff opportunities in Software Engineering, Data Engineering, Data Platforms and Data & AI worldwide. Remote is preferred; hybrid roles near Göttingen are welcome. No relocation.

morais.rm@gmail.com LinkedIn GitHub Résumé