About

I studied AI at the University of Edinburgh and CS at Riga Technical University. Since then I’ve been a programmer, software architect, machine learning engineer, consultant, startup founder.

Email: click to reveal

Areas of interest

Deep learning and its intersections with human brain, math, information retrieval, physics, biology. If a Transformer model is trained by playing a game of predict the next word, how much of human cognition can be explained by that? Can you model cognition as a set of sequence models predicting the next sound, sensation, event, visual feature, world state change? Can you model information as gas? What other physics models are applicable to neural networks?
Computing and software engineering. Taking machine learning models from Jupyter notebooks to production. Data scraping, processing, enrichment, aggregation pipelines. Pragmatic software architectures. Making software designed for change. Processes for making high-quality software products in small efficient distributed teams, without bloat and bureaucracy dragging the team down.
Entrepreneurship. In particular, the early backstage days few talk about publicly. What were the early founder decisions and circumstances that contributed to success the most? How did they get their first 100 users? What about the next 10 000? How they funded the first 365 days? What they did to achieve profitability? How and when did they hire the first 5 employees? How did they hit their first $10 000 in monthly revenue? How did they scale that to $1M? What are the consistently good and bad founding team skill mixes?
History of Silicon Valley. Many of us have heard about semiconductors, William Shockley, and Hewlett-Packard. What are the less-obvious, backstage factors that made it so fertile for inventive companies? What would it take to replicate that elsewhere and in an open-source metaverse? Which decision-making, customer development, selling, financing, legal, hiring, thinking models are consistently applicable and reusable in today’s startup companies?
Propaganda. What makes it successful? Would it become ineffective if more people learned how it works and became aware they are being manipulated by mass media, advertising, political technologists?
Anthropology, philosophy, history of science. Favorite authors: René Descartes, Jean Baudrillard, René Girard, Robert Sapolsky, Carl Sagan, Andy Clark. It took Europeans roughly 500 years to adopt Arabic numerals while just ~11 years passed from first flight to first commercial airline. It seems that the biggest obstacle to progress is legacy mindware we cling to. Are some innovations doomed to catch on for multiple generations due to some form of we always did it this way? How can the mindware be upgraded faster? Why do we generally have low awareness that much of our perceived reality is made up out of symbols and myths that have no connection to the physical reality of things?
Future of the Internet. As the saying goes, there’s no such thing as free lunch. Much of today’s Web is funded by stealing human time and attention to lure people into buying things they don’t necessarily need. Search results and social media are two prominent examples. Some of this is due to industrial overproduction and excessive economic competition. What are alternative business models that let internet companies profit while respecting human time?

Timeline

Now

A sabbatical dedicated to taking a deep dive into internals of deep learning (no pun intended).

2018 – 2022

Founder and principal developer at Le Club, an ecommerce aggregator where you could order things from abroad, even if the foreign shop didn't ship to your country. Our software scraped product listings from shops across the EU, turned raw product descriptions into structured information, translated them and made searchable through an online store-like web interface.

2014 – 2018

Co-founded and worked at various startups as a developer, founder, product manager, consultant. As a machine learning developer at ProWritingAid, I developed features used by 2M+ people willing to become better writers and presented a poster paper about the software at RuSSIR 2017. Read the paper➡

2010 – 2014

Software architect and developer at Emergn (ex-Exigen). I led development of an aggregator for the Latvian National library. It allowed readers to find books and other documents across country's libraries by aggregating and semantically enriching data from disparate sources and systems. Aggregated data was exported into Europeana via a separate OAI-PMH endpoint. I also worked on software solutions for Latvian Maritime Administration and Dictate IT (used by the British National Healthcare Service (NHS)).

2009 – 2011

In parallel to University of Edinburgh I did an MSc in logistics and e-commerce IT management at Riga Technical University IT Institute. This involved a bit of flying back-and-forth, as these were two full-time independent degrees in different countries. God bless Ryanair.

2009 – 2010

MSc Artificial Intelligence at University of Edinburgh. Here Viktor Lavrenko got me interested in search engines and I wrote a thesis on detecting web spam in search results. Also, I wrote this demo of using Reinforcement Learning to have two AI teams learn to play soccer against each other on their own, without knowing the rules: [Demo]

2007 – 2009

Software developer at Emergn (ex-Exigen). I worked on software powering the Parliament of Latvia and Latvian National Archive. There were eccentric assignments too, like making a custom Second Life island for SAP. Started as a junior and left as a senior developer ~2.5 years later.

2006 – 2009

BSc in Computer Science at Riga Technical University (RTU). In search of an open assignment topic at the library, I stumbled upon a yellow book titled "Neural Networks". That's how my fascination with AI began. My thesis was on using genetic algorithm to generate interpretable credit scoring rules.

2006

Sold first website. I physically visited local businesses asking if they needed a website until one of them said yes.

1995

Wrote first program using a listing from a programming book borrowed at a local library. I didn't have a computer at home until 2003. Luckily, our school had a computer class and the informatics teacher gladly helped. He opened class every Saturday for kids to use the internet and play games.

Credits

This website is built on top of work by:

Adam Morse and contributors behind the Tachyons CSS framework. It’s plain delightful and incredibly well-documented. Best thing since sliced bread;
Jekyll contributors, who made this lightweight static site generator;
Leon Paternoster, who integrated the two into a delightfully minimal jekyll-tachyons theme.