# Alex Varga

> ML engineer and ex-Yandex ML researcher (YSDA) building AI products from 0 to 1. Built AI agent and generative media infrastructure (DSL, model routing across 100+ models, MCP, durable execution) and a profitable bootstrapped SaaS. At Yandex, engineered the #1 ads ranking factor that lifted revenue by $6–12M.

ML Engineer experienced in AI Agent Infrastructure, Generative Media Infrastructure, Recommender Systems, DSLs, model routing and durable execution. San Francisco, CA.
Looking for: ML Engineer and Forward Deployed / Applied AI roles across AI agents, generative media, search & retrieval, and monetization ML. Based in San Francisco.

## For AI agents

- [Structured resume](https://alexvarga.xyz/resume.json): JSON Resume format

## Case studies

- **AI fitness video coaches for Simple**: Simple, a $160M ARR fitness app and category leader, wanted an ML pipeline for AI fitness video coaches. Ran R&D with my team, presented a comparison matrix of candidate models, and after they chose a custom pipeline, built and deployed it: RIFE frame interpolation plus image and video generation models. Simple launched a new app on the pipeline, and it reported higher metrics than their main app.
- **Ads recommender at Yandex**: ML research in Yandex Ads while studying at MIPT and the Yandex School of Data Analysis. Invented a new ads recommender system as my ML thesis. Increased ads revenue by 1–2% ($6–12M).
- **Hack.Moscow and 50+ hackathons**: Co-founded RH, the market leader in hackathons and ML contests in Russian-speaking countries. Ran 50+ hackathons, mostly in AI: designed tasks and evaluation for Kaggle-like competitions and led the integrations with sponsors such as NASA, Microsoft and McKinsey. 45K+ participants in total; Hack.Moscow brought participants from 52 countries.

## Experience

### Co-founder, CEO & CTO, [varg.ai](https://varg.ai) (2025-10 – present)

- One API and developer platform built for AI agents ($250K ARR): inference orchestration across 100+ frontier image, video and audio models, MCP servers, agent workflows and durable execution. Built, open-sourced and led GTM for varg.ai/sdk – a reusable DSL for multimodal generation and cloud GPU video rendering (0→1, end-to-end ownership).
- Led discovery, technical scoping and production rollout for a strategic customer automating social media at scale: hired 2 teams and shipped multi-agent workflows (data collection, analysis, creative scripting, video generation pipelines and auto-posting) producing 30,000 videos/month.
- For Simple ($160M ARR fitness app), owned end-to-end deployment of AI video coaches: ran model evaluation across candidate generative models, built a comparison matrix, and shipped a custom RIFE + image/video GPU pipeline to production – driving measurable adoption as their new app beat the main app.

### Co-founder & CTO, [VeyraX](https://github.com/VeyraX/veyrax-platform) (2025-02 – 2025-07)

- Shipped embedded AI components to production in early 2025 – live Gmail, Calendar, Slack and Jira data rendered as interactive UI inside agent workflows – ~8 months before OpenAI and Anthropic.
- Built the developer platform and SDK around it: an Artifacts-style runtime, tool-calling agents with agent memory operating 80+ connectors via UI and API, and a 3,000-server MCP marketplace behind one API key.
- Owned the integrations core end-to-end (tool-calling schema generation, OAuth integrations, MCP server deployments, Docker test infra, ~400 commits) and grew @veyraxai to ~1K followers.

### Founder, [BulkImageGeneration.com](https://bulkimagegeneration.com) (2024-07 – present)

- Built and shipped to production a 0→1 product for image generation at scale: small businesses generate up to 100 images in parallel via an AI agent workflow and inference orchestration pipeline, with programmatic SEO driving profitable growth from month 1.

### CTO & Growth (exited), [Good-Food.Ge](https://good-food.ge) (2024-07 – 2025-06)

- Joined a 10-person meal-plan delivery business as a partner with end-to-end ownership of tech and growth: hired the engineering and digital marketing teams, led technical scoping and production rollout of a custom ERP + CRM ordering platform, and ran fast experiments across acquisition and analytics – driving measurable impact (3x revenue, 30% profit margin) before selling my stake to the founder.

### Founding Engineer, [Newton (HF0, now: PolyFire)](https://polyfire.com) (2022-01 – 2022-12)

- First hire working directly with the CEO (0→1, end-to-end ownership): built a knowledge retrieval, document search and people-ranking engine aggregating data across 50+ social networks, hired and led the engineering team, and shipped to production – driving 3x organic growth to 50K users, 15 customers, 13 partnerships, #1 SEO ranking and 20% of the StarkNet ecosystem.

### Head of AI & Platform (Co-founder), [RH](https://hack.moscow) (2017-09 – 2022-02)

- To scale technical hiring and AI competitions across Eastern Europe, built a proprietary hackathon-as-a-service platform with automated ML evals and benchmarks – capturing ~45% market share across Russia, Belarus and Kazakhstan.
- Delivered 50+ hackathons and ML contests for 45K+ participants (incl. Hack.Moscow with 52 countries) across 40+ enterprise clients including NASA, Microsoft, McKinsey and Unilever, partnering with Major League Hacking.

### Head of Applied AI Lab, [Steinbeis Consulting Center AI](http://steinbeis.de) (2019-09 – 2021-04)

- Led R&D and end-to-end deployment of applied AI products (discovery → technical scoping → production rollout) at one of Central Europe's largest innovation hubs; owned B2B sales, hiring and the team. Built and rolled out an AI receipt-recognition and retail-activation platform for enterprise clients including Electrolux and Sanofi, and won industrial ML challenges for Bosch.

### Lead Engineer, Science Map, TU Berlin (2018-06 – 2019-01)

- Secured $175K in EU startup grants to build a scientific literature search and knowledge-retrieval platform: hired 9 top developers from MIPT, relocated key engineers to Germany, and shipped the MVP to production.

### ML Researcher, Yandex Ads (2017-12 – 2018-06)

- Worked across Ads Monetization and Search (BSc thesis, research → production): evaluated and ensembled 200+ production ML models and engineered the #1 ads ranking factor, lifting Yandex Ads revenue by 1–2% ($6–12M ARR) in online A/B testing.
- Built low-latency ads serving and ranking infrastructure in C and C++ for production-scale ML inference under tight latency and memory budgets.

### Data Analyst, Easy Ten (2017-07 – 2017-11)

- Built the analytics stack from scratch for hundreds of thousands of users, engineered an online experimentation and ad automation platform managing $100K/month in ad spend (running 70% of Facebook ads A/B tests), and grew an automated Instagram acquisition channel to 75K followers.

## Skills

- AI: AI agents, tool calling, MCP servers, generative image, video & audio, inference orchestration, recommender systems & ranking, RAG & search retrieval, evals & online A/B testing
- Engineering: TypeScript, Python, C, C++, Objective-C, Swift, DSLs & SDKs, model routing, durable execution, async job queues, OAuth connectors
- Leadership: 0→1 products, end-to-end deployment (discovery → scoping → rollout), B2B enterprise sales, hiring & leading teams

## Education

- Entrepreneurs First (Paris), 2025 (Selected founder)
- BSc Applied Mathematics & Computer Science, Moscow Institute of Physics and Technology (MIPT), 2014–2018 (GPA 4.75/5, math olympiad group)
- Data Science (Master's-level program), Yandex School of Data Analysis (YSDA), 2016–2018 (NLP, computer vision)

## Awards

- [MLH Top 50 Hackers of 2020](https://top.mlh.com/2020/profiles/aleksandr-malyshev)
- Parallels Scholarship (2015–2017)
- Olympiad and hackathon prizes, $20K+

## Contact

- Email: vargastartup@gmail.com
- X: https://x.com/vargastartup
- LinkedIn: https://www.linkedin.com/in/alexmsecurity