Senior Software Engineer - Market Data

Raven; Milano; On-site; EUR 120,000 - 200,000; Full time

Job summary

Raven, a proprietary crypto market-making and high-frequency trading firm, seeks engineers to own the data path from feed ingestion to research replay. You will ensure microsecond timing, decode and sequence data, and prevent gaps that cost money.

You will build and operate high-throughput, low-latency systems, work with Rust or C++, and tackle Linux networking, kernel bypass, and large-scale data quality challenges across exchanges and DeFi feeds.

Qualifications

• Built and operated high-throughput, low-latency systems where data delivery is critical.
• Experience with feed handlers, telemetry pipelines, and real-time monitoring.
• Proficiency in Rust or modern C++.
• Linux and networking depth; ability to read a packet capture and reason about kernel costs.
• Fluent English; working language across offices.

Responsibilities

• Feed handlers: ingestion from exchange venues, DeFi feeds, sequencing, gap detection and recovery.
• Latency engineering: hardware timestamps, NUMA/cache-aware layout, kernel bypass.
• Capture and replay: line-rate capture and replay for research.
• Data quality: detect staleness, cross-book issues, reorgs before traders notice.

Job description

ABOUT RAVEN

Raven is a proprietary crypto market-making and high-frequency trading firm. We trade spot, perpetuals, and prediction markets with dedicated low-latency infrastructure across eight major trading regions.

We hire competitive individuals and natural problem solvers with a record of ranking at something hard, whether competitive programming, games, or mathematics, who treat an unfamiliar problem as a challenge rather than an obstacle.

We can teach markets; we cannot teach drive.
We build our own stack end to end, from connectivity and pricing through execution, monitoring, and research data.

THE ROLE

Before a strategy can price anything, a packet has to arrive, be decoded, sequenced, book-built, normalized, timestamped, and delivered - in microseconds, correctly, around the clock, because markets never sleep. You own that path, and the capture, archive, and replay that give the same data a second life in research. A stale book or a silent gap loses money faster than almost any other bug we can write.

WHAT YOU’LL OWN

Feed handlers:Ingestion from everything we price against. Centralized exchange venues over WebSocket, FIX, and proprietary binary protocols, plus the reference feeds around them, futures and FX included - sequencing, gap detection and recovery, reconnect logic that survives the ugliest exchange behavior. DeFi feeds from the blockchain nodes around the world, where block and log streams, mempool visibility, and AMM pool state have to be reconstructed into something quotable, and reorgs, chain-tip lag, and node health are yours. Prediction markets, where books sit over discrete outcomes and the event and resolution metadata is what makes them interpretable. Work through undocumented rate limits, ghost levels, out-of-order deltas.

Latency engineering:Hardware timestamps, NUMA- and cache-aware layout, kernel bypass where it pays for itself, tail latency as the number that matters.

Capture and replay:Line-rate capture into our full-depth historical archive, and a replay path research can trust.

Data quality:Staleness, crossed books, sequence anomalies, reorgs, stale chain tips, and outcomes that stop summing to one - caught by your checks before a trader notices.

WHO WE’RE LOOKING FOR

• You have built and operated high-throughput, low-latency systems where data delivery is critical and late is the same as wrong. Feed handlers are the shortest path in, but telemetry pipelines, streaming media, and real-time monitoring build the same instincts
• The failure modes matter more than the domain. Sequencing and gap recovery, backpressure and buffering, reconnects, and telling a stale feed from a quiet one
• Deep Rust or modern C++. The stack is Rust, and you will be productive in it fast
• Linux and networking depth. You can read a packet capture, reason about what the kernel costs, tune a box, and explain a latency win end to end
• You measure before you optimize, and your numbers survive scrutiny
• Fluent English. It is the working language across both offices
• Exchange protocols (MDP3/SBE, ITCH-style, FIX), blockchain data (node operation, log and mempool streams, AMM state, reorgs), and prediction markets are a plus, not a filter. We teach markets to someone who brings the engineering
• No degree or certifications required. HFT pedigree is welcome; evidence is mandatory

LOGISTICS

• Sofia (Bulgaria) or Milan (Italy), with no preference between them. Whichever city the right person is in
• In office, not remote
• Crypto trades around the clock. You will share a sane on-call rotation with the team that owns these systems