Screenshot model distillation
A 7B-to-2B vision-model experiment, with a reproducible training and evaluation pipeline. A small proof of concept, including where the metrics misled me.
We could work together.
An early product to build, an AI hypothesis to test, or a system that needs to work for thousands of people. I bring the engineering and product decisions together.
I can turn an idea into a testable first version, work through the architecture and implementation, and help the team decide what to build next. I explain the technical trade-offs in terms of the product and the people using it.
Tell me what you’re building, where you’re stuck and what a useful result would look like.
sergey.paarfenov@gmail.comA bit about my work
I joined Yandex Praktikum as the 11th person on the team and helped build the product from its earliest stages. I worked across multiple components, designed the test architecture for its coding trainer, used by 7,000–10,000 students a day, and developed an ML model to assess student project submissions. This was before ChatGPT.
I also helped launch the QA Engineering programme in the US. At TripleTen, my work connected AI, backend engineering and the learning experience, including personalisation and the Mastery Depth Tracker.
I spent about a year developing IAWY, a startup for automated review of student project work. We launched the product in two bootcamps in Latin America.
Today I’m CTO at two projects: Aliwio, a veterinary CRM, and Symptomato, a service combining AI intake with human health specialists. I’m based in Barcelona and work across product decisions, architecture and implementation.
In the open
Agent memory & provenance
An experiment harness for measuring how memory compaction changes an agent’s decisions — including results that contradicted my own hypothesis.
Where I can help
- An AI idea worth testing
- Turn an uncertain idea into a focused experiment. Build a working prototype, test the hypothesis and use what we learn to decide what deserves a full product — before committing to a big build.
- Building an early product
- Work across the first version, architecture, backend and interface. Make decisions with the team about what to build now, what to leave out and how to take the product into real use.
- Agents you can evaluate
- Find out why an agent fails, not just whether its answer looks plausible. Build checks around tool use, memory, retrieval and recovery, with traces you can inspect and tests you can repeat.
- Smaller models, measured trade-offs
- Explore distillation and fine-tuning for a narrow task. Compare the smaller model with a baseline on quality, speed and memory before deciding whether the trade is worth it.
- Learning platforms & automated assessment
- Design test architecture for coding trainers, ML assessment of project submissions, and feedback that helps students make progress. Connect those systems to the curriculum, learner confidence and retention.
Education & courses
Stanford University
CS231n: Deep Learning for Computer Vision
Saint Petersburg State University of Finance and Economics
Computer Information Systems
The path here
Aliwio
CTO
Feb 2026 to present
I’m responsible for the technical architecture and product engineering of a CRM for independent veterinary clinics.
Symptomato
CTO
Jan 2026 to present
I lead technology at Symptomato. The service uses an AI assistant to gather context, then brings a human health specialist into the conversation.
TripleTen
AI Engineer
Dec 2021 to Jan 2026
I built AI assessment, retrieval and tutoring systems, working from architecture through production releases.
Designed and implemented a plan and execute assessment system. Reduced grading latency from 8.4 seconds to 0.6 seconds, a 14× speedup, and replaced 140 legacy text-area quizzes.
Built a Kafka based CMS vectorization pipeline processing 100,000 items a day. Delivered semantic search end to end, increasing search engagement by 9% and reducing time to content by 18%.
Built automatic practice generation that cut task authoring from 2.5 hours to 18 minutes. Shipped an internal VS Code plugin that reduced average support time per student issue by 27%.
Worked on Voice Mode for AI Tutor (Dot), connecting WebRTC, speech recognition, an LLM and speech synthesis. Helped bring it to a stable release with 97.8% successful sessions and a 1.3 second median first response.
I helped shape learning personalisation and built features that improved student retention. On the Mastery Depth Tracker, I connected product and engineering to measure learners’ confidence and identify where they needed more practice. I also helped launch the QA Engineering programme in the US.
IAWY
Co-Founder and CPO
Aug 2023 to Jul 2024
I co-founded and developed IAWY, a startup for automated review of student project submissions. I defined the product vision and roadmap for an AI tutor that integrated into existing learning platforms. We successfully launched it in two bootcamps in Latin America.
Retailhub
AI Researcher
Oct 2023 to Jan 2024
Designed a product listing enrichment pipeline: normalizing input data, extracting attributes, generating structured features and writing natural language descriptions.
Yandex Praktikum
Software Engineer
Mar 2018 to Dec 2021
I was the 11th person to join the team and one of the people who built the product from its earliest stages, contributing across multiple components.
I designed the test architecture for the coding trainer used by 7,000–10,000 students a day.
I developed an ML model for assessing student project submissions before ChatGPT, and contributed to the launch of the Web Developer programme.
Thingyfy
Software Engineer
Oct 2019 to Oct 2020
Built startup products with React and TypeScript, focusing on adaptability and scalability.
Yandex School of Data Analysis
Learning Specialist
Oct 2018 to Jan 2020
Worked as a learning specialist in Moscow.
Sravni.ru
Software Engineer
Nov 2017 to Oct 2018
Built insurance products in React.
Earlier work
Engineer and teacher
2016 to 2019
European Gymnasium
Computer science and web development.
Spaces
Commercial property product engineering.
MCS
Arduino, Swift, Python, web, and VR teaching.