S. BECONIS

About

I'm Saulius Beconis, an applied AI engineer and software builder in Lithuania. I build systems for specific, messy problems, the kind that don't fit neatly into an off-the-shelf product.

S. BECONIS is my one-person engineering practice. There is no team behind the name: you work directly with the person who designs and builds the system.

What I work on

  • Retrieval and knowledge systems: search and question answering over an organisation’s own documents, measured rather than guessed.
  • LLM integrations and agentic workflows: AI inside existing software, with tools, permissions and checks.
  • Document and data pipelines: structured, validated data out of messy sources.
  • Local and private AI: models that run on your own hardware when data can’t leave it.
  • Generative media infrastructure: image and video pipelines built as dependable systems.
  • Custom software built around real operational needs, with or without AI.

How I think about it

AI is usually only one part of the system. Sometimes the hard problem is retrieval. Sometimes it’s data quality, orchestration, evaluation, interfaces, latency, deployment, or simply writing good software. I don’t believe every problem needs a language model, and I’m happy to use conventional engineering when it’s the better solution.

I use modern coding agents heavily, and they make me faster. They don’t replace engineering judgement: architecture, verification, testing and understanding the system still matter. This site was built the same way, and its Lab shows how I measure the systems I build.

Background

I started in 2019 with hidden Markov models for genetic sequence analysis, in university research and the iGEM synthetic biology competition. That led into computational biology and bioinformatics, and from there deeper into applied AI, local models, agentic systems and custom software. My work still sits where software, machine learning, scientific computing and biology meet.

Much of my current work is experimental: building and evaluating AI systems, testing what new models can actually do, developing local inference workflows, working with multimodal and generative pipelines, and applying machine learning to scientific and acoustic data. The results I can share end up in the Lab and in my writing.

We’ll probably work well together if

  • your problem doesn’t fit off-the-shelf software: an unusual dataset, a difficult workflow, a process too specialised for generic tools, or a system that doesn’t exist yet;
  • you want a straight answer on whether AI is the right tool, including when it isn’t;
  • you’d rather talk to the engineer who builds it than to a sales team;
  • you care about evidence: a working prototype on your data and measured results before a bigger commitment.

If the problem is interesting and technically solvable, there’s a good chance I’ll want to build it. Start a project or write to saulius@sbeconis.com.