Marcin Jakubowski · Principal Data & AI Engineer

Enterprise AI & data platforms, engineered at the standard the top demands.

I'm a one-person consultancy behind mission-critical AI platforms, cloud data pipelines and executive analytics for some of the world's most demanding institutions — global investment banking, healthcare and diagnostics, international finance. Principal-level engineering, no hand-offs, no juniors.

Start a conversation Explore the work

AVAILABLE FROM JANUARY 2027 · B2B CONTRACT · REMOTE · PRINCIPAL / STAFF IC

0+
years engineering data systems
0%
processing time cut in flagship platform migration
0+
AI services unified into one enterprise platform
0%
CRM adoption growth driven by my analytics

Proven in high-stakes, regulated environments

Global investment bank Global healthcare & diagnostics leader Global banking group European automotive finance Global industrial technology group
Marcin Jakubowski — portrait
Marcin Jakubowski
Founder · Principal Engineer & Architect
Poland · working worldwide, remote-first
BSc Applied Physics, Warsaw University of Technology
Linux native since 1997 · polyglot programmer
The person behind the work

Twenty years of engineering, from server rooms to agentic AI.

I work alone by design. When you engage me, you get Marcin Jakubowski — a principal-level engineer and architect who has spent two decades building the systems that businesses actually run on, most recently as the architect and lead engineer of an enterprise AI platform at a global investment bank, under some of the strictest regulatory and security constraints in finance.

The career arc runs from the metal up: hardware, networks and infrastructure first; then databases, ETL and BI at global scale; then cloud data platforms; and now the frontier of applied AI — agents, RAG, evaluation systems, LLM platform engineering. That full-stack depth means I debug the whole system, not just my layer of it.

Renewed beyond standard limits

Contract renewed repeatedly beyond standard tenure caps — each time at the client's initiative — because the platform knowledge was considered irreplaceable.

Speaks business, writes code

Trusted in direct contact with managing directors and C-level stakeholders — translating business intent into architecture, and shipping it personally.

Wired for extreme pace

Years in environments where leadership requests demoed on Monday ship by Wednesday. AI-augmented engineering workflow, senior judgment on every line.

Built for the love of it

Long before the enterprise work, I designed, built and shipped Żulionerzy — a cult-classic Polish mobile quiz game with 7M+ downloads in a country of 38M, entirely solo. It's the same reason I still write code personally today: I do this because I love it, not because the invoice clears.

What I do

Four disciplines. One standard: production-grade.

I don't do proofs-of-concept that die in a slide deck. Everything I build is designed to run in production, at enterprise scale, under real regulatory and security constraints.

Enterprise AI Platforms

My flagship discipline. I architected and built the division-wide AI platform of a global investment bank — from a single RAG application to a full multi-agent ecosystem serving thousands of bankers.

  • Agent orchestration, routing & fallbacks (LangGraph)
  • RAG with claim-level source verification — every statement traceable
  • Unified LLM gateways across a dozen upstream services
  • Auto-evaluation pipelines, model benchmarking, LLM ops
  • Skill & tool execution engines, MCP integration

Data Engineering & Cloud ETL

Petabyte-scale pipelines, cloud-native and fully automated. I've migrated entire ETL portfolios off legacy platforms — solo, on time, and reliable enough to terminate the old licences early.

  • AWS-native pipelines: Glue, Lambda, Step Functions, EMR
  • Snowflake, dbt & data mesh / data product architecture
  • Full infrastructure-as-code: Terraform, CDK, GitLab CI/CD
  • Legacy ETL migration (Informatica, SSIS, on-prem SQL)
  • Data quality, lineage, access policies & governance

Business Intelligence & Analytics

Dashboards that executives actually use. I build purpose-designed warehouses and reporting suites that replace days of manual work with a one-minute self-service experience.

  • Executive & C-level analytics (Tableau and beyond)
  • Purpose-built data warehouses & semantic layers
  • KPI frameworks: adoption, revenue, coverage, productivity
  • Row-level security across roles, regions & segments
  • Automated report generation replacing manual decks

Architecture & Technical Leadership

Fluent in both engineering and business. I've led teams of 40+ developers through migrations that other vendors failed at — and left every organization with a stronger engineering culture.

  • Solution & platform architecture for regulated industries
  • Team leadership, mentoring & managed delivery
  • CI/CD & DevOps transformation of legacy platforms
  • Stakeholder management up to C-level and MD-level
  • Rescue engagements for stalled, high-stakes projects
0bn+
rows processed in a single reporting platform
$0k+
annual licence savings from one migration alone
0+
production dashboards designed & delivered
0+
engineers led across global delivery teams
Selected work

Engagements that moved the needle — measurably.

A sample of what I've architected and delivered. Client identities are withheld out of respect for confidentiality; the engineering and the outcomes are real.

Global investment bank · AI platform

The AI platform an entire banking division runs on

A dozen AI applications, each with its own frontend and protocol, consolidated into a single platform serving thousands of bankers. I designed and built the unified gateway: one streaming protocol across all upstream services, model routing with prioritized fallbacks, and claim-level source verification — every AI statement traceable to the exact passage that produced it.

From there the platform grew into a full agentic ecosystem: LangGraph orchestration with fast/deep-research modes, personalization, graph-based knowledge retrieval, an automated evaluation pipeline that benchmarks every new foundation model, a skill execution engine over hundreds of MCP tools, and full cross-service observability.

12+upstream AI services unified behind one protocol
100%of AI claims traceable to verified sources
1000sof daily users across a global division
<48htypical idea-to-production turnaround
Investment banking · GenAI

Earnings intelligence with verifiable answers

A RAG application over tens of thousands of earnings-call transcripts: chat, cross-period comparison, topic alerts and auto-summaries — with a claim verification engine that pins every answer to the exact transcript passage. It became the seed of the division-wide AI platform.

10k+
transcripts covered, growing quarterly
80%
of backend engineered by me personally
Investment banking · Cloud ETL

Legacy ETL portfolio to cloud-native AWS — solo

Complete migration of a division's CRM data flows from Informatica to AWS Glue, Step Functions and Lambda, fully Terraform-managed — delivered single-handedly, ahead of deadline, and so reliable the legacy licence was retired early rather than renewed.

$100k+
annual licence cost eliminated
100%
cloud-native, infrastructure-as-code
Global healthcare · Data mesh

12 hours to 25 minutes on Snowflake

Re-architecture of a business-critical CRM reporting platform from an overgrown SQL Server estate to a lightweight Snowflake data-product architecture with dbt. Became the poster-child migration for the entire organization — a parallel vendor rewrite of the same system ran 20× slower on hardware twice the size.

−96%
processing time
20×
faster than the competing rewrite
Investment banking · Data science

Rescuing a 250k-record contact acquisition

An expensive third-party contact dataset was 70% unusable after manual review. I built an 11-algorithm entity-matching engine with confidence-weighted consensus — pattern matching, ML and massive SQL over DuckDB — turning a write-off into one of the division's richest CRM assets.

30→80%
usable records
95%
C-suite contact coverage achieved
Investment banking · Executive BI

Monthly exec decks: from days to one minute

A purpose-built warehouse and Tableau suite generating full presentation-grade reporting — revenue streams, deals, coverage KPIs across regions and industries — on demand. Work that took senior bankers days each month now takes under a minute; a companion adoption dashboard drove CRM usage up 300% within a quarter.

+300%
CRM adoption in 3 months
days→1min
reporting turnaround
Global healthcare · DevOps

From 2 releases a year to release-on-demand

Designed and drove the CI/CD transformation of an entire BI platform: TFS-to-Git migration, automated unit testing, one-click multi-environment deployments — including a fully automated Tableau deployment pipeline. Adopted across 20+ projects and 100+ database developers.

On-demand
releases — up from 2 blackout releases/year
100+
developers on the new toolchain

Engagements are described generically to honor confidentiality agreements. Employers and clients are named in my CV and in conversation; this page stays deliberately generic.

Technology

The stack I'm dangerous with.

Tools change; fundamentals don't. This is what I currently ship to production — behind it sit twenty years of SQL, distributed systems and hard-won operational judgment.

AI & LLM Engineering

LangGraphLangChainRAGClaim verificationAgent orchestrationMCPAuto-evalsClaude · GPT · GeminiFastAPIOpenSearchAWS Neptune

Cloud & Infrastructure

AWS GlueLambdaStep FunctionsECS / FargateEMRS3 · RDS · DynamoDBTerraformCDKKubernetesGitLab CI/CD

Data Platforms

SnowflakedbtApache SparkSQL ServerPostgreSQLMongoDBDuckDBSalesforceSAP CO-PAData mesh

Languages & BI

PythonSQL (expert)JavaC#TypeScriptTableauPySparktSQLtAI-assisted engineering
How I engage

Small footprint. Outsized outcomes.

01 / UNDERSTAND

I learn your business first

Investment banking, diagnostics, finance — I've become fluent in each. I start by understanding the process and the P&L behind your data, not by proposing architecture diagrams.

02 / ARCHITECT

I design for production, day one

Security, governance, IaC, observability and cost are part of the first design, not an afterthought. Everything I ship is built to survive audits, scale and staff turnover.

03 / DELIVER

I build it — personally

No delivery pyramid, no junior rotation. The person who designed your system writes the critical code, mentors your team, and stays accountable for it running in production.

Work with me

Have a platform to build — or one that needs rescuing?

I take on a small number of engagements at a time, so every client gets principal-level attention. If your data or AI initiative is too important to hand to a delivery pyramid, let's talk. Currently completing a multi-year engagement at a global investment bank; booking conversations now for a January 2027 start.

Start a conversation
marcin@hqdev.net
CET · full overlap with UK & EU business hours
Remote-first, worldwide
Available for new engagements from January 2027
LinkedIn

References available on request — including senior engineering leadership at a global investment bank.