Tom Smykowski is a senior embeddings search engineer and product-minded technical leader with nearly two decades of experience building and scaling software products. For teams in the EU, UK, and US seeking to implement robust, production-grade semantic search, retrieval-augmented generation (RAG), or vector database solutions, Tom provides the rare combination of deep technical expertise in embeddings and vector search, full-stack engineering prowess, and a product-focused mindset that ensures systems deliver real user value. He specializes in taking ownership from initial architecture and prototyping through to deployment, monitoring, and optimization, making him an ideal partner for startups and established companies alike.
From Concept to Production-Ready Semantic Search
Implementing embeddings-based search is more than just plugging in a vector database. It requires a holistic understanding of data pipelines, embedding model selection, indexing strategies, and result ranking to create a system that is both accurate and performant under load. Tom approaches each project with this full lifecycle view. His work as a Staff Software Engineer at Croissant, a workplace platform, involved architecting complex, compliant systems for collaboration, a discipline that translates directly into building reliable, scalable search infrastructures where data integrity and performance are paramount.
His experience extends to high-traffic environments, having led frontend and design system work at Printify, a platform serving over 10 million merchants. This exposure to systems at scale informs his embeddings search engineering, ensuring that vector search implementations are designed with efficiency and maintainability from the start. He focuses on creating clean abstractions, comprehensive testing, and observable systems so that your semantic search capabilities can evolve smoothly with your product.
Architecting Scalable RAG and Vector Search Systems
The architecture of an embeddings search system dictates its success. Tom specializes in designing solutions that balance accuracy, latency, and cost. This involves critical decisions around embedding models, the choice of vector database (whether dedicated like Pinecone or Weaviate, or PostgreSQL with pgvector), chunking strategies for documents, and hybrid search approaches that combine semantic and keyword-based techniques. His full-stack expertise with Vue.js, TypeScript, and Supabase ensures he can build the entire pipeline, from the backend APIs that generate and store embeddings to the frontend interfaces that deliver a seamless search experience.
For projects requiring retrieval-augmented generation (RAG), Tom engineers the retrieval component to be a robust, low-latency service that provides high-quality context to LLMs. He understands that the quality of the RAG output is directly tied to the precision of the semantic search. His work emphasizes not just initial implementation but also the ongoing evaluation of retrieval performance, using metrics like hit rate and context relevance to iteratively improve the system.
Engineering for Performance and Real-World Data
A common pitfall in embeddings search is neglecting performance as the vector index grows. Tom engineers systems with scalability in mind, implementing strategies for efficient indexing, query optimization, and caching. He has practical experience with the challenges of real-world, messy data from his tenure at industrial SaaS and ERP companies like Fream and Insoft Serwis, where he built data-intensive applications. This background is invaluable for designing data preprocessing and normalization pipelines that ensure your source data is optimally prepared for embedding generation.
Furthermore, his product thinking ensures the search system aligns with business goals. He focuses on user-centric metrics beyond pure technical performance, such as query conversion rates or user satisfaction scores. This approach means the embeddings search system he builds doesn't just work technically, it actively contributes to key product outcomes, whether that's helping users find information faster, improving customer support, or powering personalized recommendations.
Integration and AI-Native Development Workflows
Tom doesn't work in a vacuum. He excels at integrating embeddings search capabilities into existing product ecosystems and development workflows. As an advocate for AI-native software development, he can guide teams in adopting tools and practices that leverage embeddings search effectively. This includes setting up AI-assisted code review systems, implementing AI quality gates, and automating testing pipelines where semantic understanding of code or content is required, all areas where his related expertise in AI testing automation and AI-assisted code review directly applies.
His remote collaboration style, honed through years of working with international teams from his base in Poland, is built on clear communication and async-friendly processes. He seamlessly overlaps with EU, UK, and US East Coast time zones, ensuring smooth integration with your engineering and product teams. He is adept at using diagrams, technical documentation, and collaborative tools to articulate complex architectural decisions around search systems to both technical and non-technical stakeholders.
Proven Delivery for Marquee Clients
Tom's credibility is backed by a track record of delivery for notable organizations. Beyond his staff engineering role at Croissant and scaling work at Printify, he has served clients like DeLaval, GE Healthcare, Unilever, and the Ordnance Survey. As the founder of Websoul, he scaled a marketing automation platform to over 32,000 businesses, an experience that underscores his ability to own and grow a complex SaaS product, a mindset he brings to every embeddings search project. These engagements demonstrate his capacity to navigate diverse industry domains and technical challenges, building systems that are reliable and deliver value.
His background also includes significant work in the public and academic sectors, such as projects for the Polish Tourism Organisation, the Institute of National Remembrance, and the Silesian University of Technology. These projects often involved managing large, unstructured datasets and presenting them through intuitive interfaces, a foundational skill for anyone working with embeddings and search. This diverse experience means he can quickly adapt to your specific domain, whether it's e-commerce, industrial IoT, healthcare, or digital humanities.
When This Is a Good Fit
Engaging Tom as your embeddings search engineer is an excellent fit if you are a founder, CTO, or technical lead needing to build or scale a semantic search, RAG, or vector database system and want a senior engineer who can own the problem end-to-end. It's ideal if you value product-minded technical leadership, clear communication with remote and cross-time-zone teams, and a focus on creating maintainable, scalable systems that directly impact your users.