Szczepan Gurgul
Software & Blockchain Engineer
Software engineer with blockchain expertise and a background in distributed systems. I build secure, maintainable systems and tooling - from smart contracts and DeFi primitives (DEX/AMM) to performance-oriented infrastructure. Currently working as an AI × Blockchain Engineer @ Nextrope, building AI agent workflows for smart contract security analysis and EVM bytecode auditing.
Education
M.Sc. Informatics, University of Zürich — 2026 (magna cum laude)
B.Sc. Mechatronic Engineering, AGH University of Science and Technology — 2023
Stack
AI
Blockchain
Software Engineering
Programming Languages
Achievements
BCOLN Challenge Award - Best Distributed dApp
Won the BCOLN Challenge Hackathon (2024) with a Decentralized Sports Betting Platform, competing against 10 teams from across Zurich.
Uniswap v4 Hook Incubator
Selected for the Uniswap v4 Hooks Incubator (Atrium Academy), protocol engineering cohort focused on designing, implementing, and testing production-ready Uniswap v4 Hooks. Cohort starts in October 2026.
Conference Presentations
- Mumbai, India 2025 - AI Agents for MEV on Rollups (BCK25)
- Zürich, Switzerland 2024 - Scaling DeFi with ZK Rollups (BCK24)
Publications
AI Agents for MEV Extraction on Rollups
The Physical Society of Japan (under review) · 2026
- Built a research MEV bot, running against live mainnet liquidity pools for EVM rollups.
- Used deterministic economics with CPMM/CLMM formulas to calculate MAV. Created stochastic economics to model sequencer/mempool uncertainty.
- Implemented and compared MEV strategies: cross-DEX arbitrage, sequencer spamming, sandwich attacks
- Run a multi-agent setup to orchestrate the MEV bot and gather market insights
- Benchmarked PnL, execution cost, and risk under realistic L2 constraints (fees, price impact, ordering)
arXiv (preprint) · 2025
- Deployed a private ZK rollup including sequencer and prover on zkSync (Elastic Chain stack)
- Forked Uniswap V2 as the AMM under test
- Built a load balancer / tx generator to stress-test throughput
- Benchmarked SNARK proofs, TPS (74 vs ~15-20 on mainnet), soft & hard finalities, and gas patterns