Maciej Piekarski

Working on a greenfield project: the next generation of parking lot monitoring devices with capabilities beyond license plate recognition. Think cameras with AI on the edge, feeding into more powerful vision-language models in the cloud - with the goal of converting advances in AI into tangible benefits for parking lot owners and managers.

Together with a small army of robots (mostly Claudes), I am responsible for infrastructure, CI/CD, backend, frontend, communication with IoT devices - and anything else required to solve user problems as fast as possible. Beyond my own project, I share what I learn about working with agents - from reading and experimentation - with other people and teams, helping them become more productive with these tools too.

• Set the project up from scratch, converting a scrappy PoC (infra clicked out in the AWS console, code modified directly on lambdas) into a production-ready system. This involved defining infrastructure as code with Terraform, building the serverless backend and frontend, configuring Auth0 for authentication, and creating a full stack aligned with company standards that gives a lot of flexibility to iterate.
• Revisited the CI pipeline to cut unnecessary complexity, rebuilt it for my project, and advocate for other teams to adopt a "you build it, you run it" DevOps mindset.
• Introduced Claude into a production image analysis pipeline for anomaly detection. Built a testbench to provide guardrails and tune the non-deterministic system for consistent production behaviour. Pre-processing with Ultralytics YOLO to provide guidance, experimenting with various approaches and models.
• Evaluated and deployed Ultralytics YOLO and Meta SAM models, figuring out how to run them cost-effectively in a way that fits the project's existing serverless infrastructure.
• Designing the architecture for the next-generation device platform - edge AI processing combined with cloud-based VLMs for more complex analysis.

Tech: React, Python, AWS Lambda, Terraform, Ultralytics YOLO, Meta SAM, Claude API

Travelled, rode bikes, met great people.

Created, from idea through launch to monetisation with hundreds of subscriptions, two innovative products in the home energy management space: tado Balance and Balance for Heat Pumps. The first product was built by a team of six in two months.

Directly responsible for all engineering aspects - backend and frontend - including choosing technologies, integrating the new product into the existing system landscape, and monetising with subscriptions.

Product & Engineering

• Gradually increased product reliability by improving testing, monitoring, and alerting to match the product's growing impact on users and higher expectations for paid services.
• Built subscription management and integrated with app stores.
• Designed APIs and internal system architecture. Prepared and managed migrations (Elasticsearch, Redis, MySQL).
• Made architectural decisions to fit the new product into the existing landscape from both UX and technical perspectives.
• Participated in the redesign and refactoring of core backend systems to improve platform maintainability for the new hardware generation.

Product Discovery & User Research

• Drove 10+ customer interviews, did a field visit shadowing HVAC installers at a partner's HQ to understand their workflows and needs.
• Contributed to product and UX decisions by helping analyse the market, understand users, identify opportunities, and help build and maintain the impact/effort matrix.

Cross-functional Work

• Helped the marketing team craft press releases and prepare presentations. Championed an "underpromise, overdeliver" approach.
• Successfully advocated for replacing legacy hardware by quantifying long-term engineering and customer support costs and building a compelling case for a one-time investment.
• Built an AI pipeline using Whisper and GPT to automate transcription and summarisation of customer interviews, saving significant time per interview.
• Member of the sustainability team - calculating company emissions, working with logistics to reduce packaging size and materials.

Knowledge Sharing & Hiring

• Weekly product demos, cross-team knowledge-sharing sessions, brown bag sessions, contributors' meetings (see How I Work With People above).
• Conducted technical interviews and experience days. Helped craft the interview guide/handbook.

Tech: Angular 19 (declarative/reactive with signals), TypeScript, AWS API Gateway, Lambda, DynamoDB, SQS, SNS, SES, CloudFormation, CloudWatch, GitHub Actions

Owned a critical part of a cloud-based IoT platform serving millions of connected devices - responsible for its reliability, scaling, and evolution from tens of thousands to millions of connections. This meant understanding the system deeply enough to make architectural decisions about data pipelines, caching layers, and database scaling under real production pressure.

• Significantly improved system efficiency and reduced database and server costs by optimising critical device data processing code. Investigated performance issues in Spring Boot servers and identified bottlenecks in RDS, deciding on scaling approaches.
• Shaped internal system architecture and designed internal and external APIs.
• Ensured high code quality through rigorous reviews, robust test coverage, and establishing team coding standards.
• Enhanced system reliability by refining monitoring and alerting. Successfully managed critical data store migrations (Elasticsearch, MySQL) with minimal user impact.
• 2 years of on-call duty for critical infrastructure, combined with 3rd level support for the CX team - figuring out what is a bug vs. a feature, writing issues for fixes and features, updating the status page, writing post-mortems, presenting learnings at tech demos, helping build monitoring and alerting, deciding on important metrics, and preparing emergency checklists.
• Managed critical production incidents including customer communication, root cause investigation, and developing operational checklists to prevent recurrence.
• Improved deployment velocity and reliability by speeding up test execution and migrating CI/CD pipelines to GitHub Actions.
• Provided technical analysis and implementation plans for new use cases and maintenance efforts.

Tech: Java, Kotlin, Groovy, Spring Boot, Redis, Amazon Kinesis, RDS, InfluxDB, Jenkins, Datadog

Worked closely with domain experts from logistics and finance to develop solutions for optimising truck routes and streamlining accounting processes. Contributed to a new ledger project.

• Focus on understanding core requirements and driving iterative approaches to develop effective, user-focused solutions.
• Organised company-wide brown bag sessions to foster technical discussions and share knowledge across teams and departments.

Tech: Java, Spring Boot, AWS (RDS, SQS, SNS, Kinesis), Terraform, Mesos, Grafana, Kibana

Contributed to creating and improving a critical application ensuring regulatory compliance for UBS.

• Emphasised clean coding practices and disciplined test-driven development approach.
• Participated in rigorous code reviews.
• Worked with strict availability and performance requirements.

Tech: Java, Spring

Contributed to building a highly available and dynamically scalable IAM solution for a network management system.

• Acted as Scrum Master, driving various team and department meetings including for geographically distributed teams.
• Travelled to Ericsson's site in Ireland to interface with teams there, learn more about the product, and solve issues ad hoc.
• Onboarded new joiners to the business domain and technical details of the system.

Tech: Java (Ericsson frameworks and internal tooling)

Created innovative and highly accurate measurement devices for materials science laboratories. The devices measured the Seebeck effect coefficient distribution across material samples, using image processing to assess sample uniformity and an automated test stand for physical property measurement. This doubled as my MSc thesis work - combining image processing with physical measurement automation.

• Defined device parameters based on market demand to ensure product-market fit.
• Marketed the product at an industry conference to find first customers.

Implemented software tools for early diagnosis of pulmonary hypertension through 4D flow MRI analysis - detecting vortices in blood flow patterns to identify arterial hypertension.

Designed and implemented software for automated test stands used for QA of hardware components.

Thesis completed in collaboration with Tegeos: image processing for detecting material sample uniformity (Seebeck effect coefficient distribution) and building an automated test stand for physical property measurement.

Thesis on visualising protein folding.

Image processing, FPGAs, control systems.