01
AI that works
From strategy and prototype to a reliable production system: LLMs, search, image analysis and intelligent workflows.
AI systems · Data · Cloud
I turn ambitious ideas into AI that works in the real world.
A technology partner for teams that want to build smarter: from AI and data strategy, through a prototype, to a production-ready system.
Start with the problem ↘20+years of experience
AI + Datafrom prototype to production
GCP & AWScloud production
PythonGCP & AWSLLM systemsComputer VisionData Engineering
Project experience with
What I help with
Technology should take weight off your team.
I work where product, engineering and measurable outcomes meet.
02
Data that creates an edge
I design data pipelines and models that replace gut feeling with better decisions.
03
Cloud ready to grow
Scalable backends, APIs and event-driven architecture on GCP and AWS — built to be maintained.
More than a portfolio
Experience that informs the next move.
I have over 20 years of practice building products and systems. Some projects are confidential — so instead of a full list, here is the kind of problems I deliver.
AI & media intelligence
Video and image analysis workflows with Vertex AI, Vision AI, LLMs, Cloud Tasks, BigQuery and Datastore.
How it works
- Video & images
- Cloud Tasks
- Vertex AI · Vision AI · LLM
- BigQuery, Datastore
- Results
Data platforms
Data pipelines in GCP, live analytics (Pub/Sub, Dataflow), automation and solutions that work beyond a demo.
How it works
- Data
- Data pipeline · GCP, Airflow · Cloud Functions; Live analytics, Pub/Sub · Dataflow
- Automation
- Results
Retail & predictions
AI supporting merchandising and product comparison at e-commerce and retail scale.
How it works
- E-commerce & retail data, ETL · GCP
- Cascading T-SQL procedures
- AI
- Stock allocation forecasting, Visual product comparison
- Results
Risk & trust systems
Compliance systems, anomaly detection, classification and text analysis.
How it works
- Text analysis, Kafka · Redis
- Customer classification, Fraud & anomaly detection, Spam filtering, Python · SQL
- Decision tree
- Actions
- Metrics
Project experience
Experience with systems deployed to production and still running today.
I have worked for organisations where data quality, reliability and delivery speed directly affect the business.
For Google: a system that evaluates visual ad creatives — video and image analysis with Vertex AI, Vision AI, LLMs and Vertex Search, workflows on Cloud Tasks, BigQuery and Datastore.
UK state-owned energy companies: an energy source broker on AWS IoT and Lambda. The Football Association: data pipelines in GCP with Airflow, Cloud Functions, Pub/Sub and Dataflow. AWS Connect integration.
AI for stock allocation forecasting across online and physical stores, visual product comparison, ETL on GCP and cascading T-SQL procedures. In the payments team: anomaly detection in transactions.
Hydra team: compliance systems in production — fraud detection, customer classification, spam filtering and text analysis (Python, SQL, Kafka, Redis).
Eight years building the Gandalf.com.pl bookstore: from a team of five to Poland’s leading book retailer. Shop, admin panels, SEO and campaigns.
From media and image analysis, through cloud and data, to e-commerce, compliance and predictive systems. See the full background on LinkedIn ↗
Who is behind it
Bolesław Badziąg
Technology partner · AI / Data / Cloud
Smart Choice is my company — you talk directly to the person who designs and delivers the solution. For more than 20 years I have been building products and systems: from the Gandalf.com.pl bookstore, through compliance at GetResponse and predictions at LPP, to video and image analysis with Vertex AI for Google.
LinkedIn profile ↗AI-native delivery
Faster to a working prototype. No shortcuts in engineering.
I work with the new generation of AI tools, so a working prototype often takes weeks rather than months. The rigour stays the same: unit, integration and load tests, code review, sound architecture and security from the first commit.
Unit testsIntegration testsLoad testsCode reviewSecure architecture
AI agents and models: Claude, Cursor, Codex, Grok, Antigravity — chosen for the job.
Collaboration without magic
From the first conversation to results.
A short audit, an intensive proof of concept or longer engineering collaboration — all are possible.
- 01
We understand
Goal, users, data, constraints and the success criterion.
- 02
We validate
The smallest sensible experiment before you invest in a full solution.
- 03
We deliver
Architecture, implementation, tests and a plan for what comes next.
Three ways to start
Timing is indicative — we agree the scope after the first conversation.
approx. 1–2 weeks
AI & data audit
I review your data, processes and whether AI makes sense here.
- Report on your data and processes
- Target architecture
- A clear recommendation: yes or no — and why
approx. 2–4 weeks
PoC / MVP sprint
We test the hypothesis on your real data, in an isolated environment.
- A working prototype on real data
- A success metric agreed before we start
- Conclusions: build on it, change course or stop
long-term collaboration
Production delivery / Fractional AI Architect
I develop the system with you or support your team as an architect.
- Production-ready architecture, implementation and tests
- Code review and architecture support for your team
- Knowledge transfer — the system keeps working after we finish
Who I am not for
Not the right fit if you…
- are looking to implement AI without a defined business problem,
- only need a presentation or demo, with no plan for ongoing maintenance,
- expect confirmation of a decision already made rather than an independent assessment,
- plan a full rollout straight away, without a stage to validate the hypothesis.
A good fit if you…
- have a specific process that takes too much time or loses information,
- need a production-ready solution — with tests and a plan for further development,
- expect a sound recommendation — including a negative one — before deciding to invest,
- want to start with the smallest meaningful experiment on real data, with a success metric agreed before the start.
FAQ
Questions before the first conversation
Do you work alone or join our team?
Both. Smart Choice is my company, so you talk directly to the person who designs and delivers the solution. I can run the project on my own or join your team and support it with architecture, code review and knowledge transfer.
How do you handle data security and privacy? Is our data used to train public models?
Client data stays in the client’s own infrastructure or cloud account wherever possible. I use model services in modes that do not use data for training. We agree data access rules at the start, and a PoC runs in an isolated environment.
How do we settle payment?
We agree it individually: a fixed price for an audit or PoC, or time-based billing for longer collaboration. The initial assessment of your problem before we start is free.
What if the audit shows that AI does not make sense?
That is a good outcome too. You get an honest recommendation — including “no” — and pointers on what is worth checking or improving instead, before you invest in a costly project.
How soon can we start?
Describe the problem in the form below — I will reply with a proposed first step. We then agree the start of the audit or PoC together, depending on scope and data availability.
Free initial assessment
Do you have a process, data or product that should work smarter?
Describe the problem in a few sentences. You will get an honest answer on whether AI makes sense here — and where to start.
Not ready to talk yet? See 10 questions before starting an AI PoC →
Smart ChoiceBolesław Badziąg ul. Folwarczna 23B
84-207 Bojano
Poland NIP: 9581251173 +48 690 760 137 kontakt@smartchoice.dev