Roman Vasetskiy
About
I build high-performance backend systems on the JVM — mostly in Scala and Java, sometimes Kotlin. For the past 20+ years I have worked in financial infrastructure: trading platforms, treasury risk, market-data pipelines, ETL and batch-grid compute at scale.
My focus is on the part of the stack that is boring until it breaks: latency, throughput, correctness under load. I have spent many afternoons staring at flame graphs, cache-hit-ratio dashboards and grid-timeout logs, and a surprising number of production wins turned out to be a combination of three things: removing something unnecessary, moving a computation closer to the data, and being honest about what you don't need to recompute.
Currently
Leading architecture and delivery for four Treasury services (Risk / Funding / Controls) with a small team of engineers; end-to-end ownership across design, implementation and cross-team coordination.
- Cut EOD Treasury Risk/Funding runtime on a distributed batch grid from P95 5–8h to P95 10–15m over ~400k positions, reliably completing within the post-trade window.
- Eliminated grid node-timeout failures in a critical risk scenario by redesigning caching and data access, guided by a graph-based profiler (hotspots, cache hit/miss, CPU/memory).
- Built a reusable validation + health-check framework adopted by 4 teams (7–30+ engineers), covering 13 critical processes and reducing investigations from 1–2/week to near-zero.
Previously
Owned a trading data integration platform feeding Salesforce and analytics from 50+ Oracle databases. Managed a distributed team of 24 engineers with full remote leadership and on-site visits.
- Reduced Salesforce data latency from 5–6 hours to ~3–5 minutes, processing ~1M events and ~500k trades/day.
- Designed SLA-tiered delivery with business and engineering stakeholders — event-driven updates for trades, timer/EOD for lower-priority outputs.
- Built an in-house Kotlin/Spring ETL engine replacing third-party tooling; achieved ~$900k/year in savings after ~1 year of hardening and scaling.
Joined as a backend engineer, progressed to Tech Lead in a 15-person startup with a flat structure (Tech Lead → Director → Owner). Built reconciliation and background reporting systems on Java/Oracle.
Stack
Languages: Scala, Java, Kotlin, SQL/PLSQL
JVM & Frameworks: Akka, Spring, Apache Spark, Cats Effect, ZIO (hobby)
Storage: PostgreSQL, Oracle, MongoDB
Infra & Observability: Jenkins, Grafana, Splunk, profiling tooling, distributed batch grid
Systems I like building: batch/streaming hybrids, caching layers, validation frameworks, grid scheduling