00 — crypto
Crypto infra for the desks that take it seriously. The other ninety percent of DeFi we leave alone.
For market-making desks, OTC desks, and crypto-native funds measuring the work in fills, not retweets.
AI in crypto, real share
~10%
we do that part, skip the rest
reorg-aware indexers
4-block
confirmation window
engagement shape
infra only
alphas stay yours
01 — what we do
Concrete deliverables. Named tools. Honest constraints.
01
Market-maker tooling.
Quote engines, inventory management, and skew logic tuned to your flow. Latency budgets in mono, not in a deck.
02
On-chain data pipelines.
Indexers across the chains you actually trade. Reorg-aware, backfill-safe, with a clean schema your analysts can query.
03
OTC ops infra.
RFQ workflows, settlement tracking, and counterparty exposure dashboards. Built to survive an audit and a 3am incident.
04
MEV-aware routing.
Order routing that prices in protection, private orderflow, and bundle inclusion probability. We use a model only when the eval beats a tuned heuristic.
05
Monitoring and dashboards.
Real-time PnL, inventory drift, and on-chain anomaly alerts. Pages someone, not a Slack channel everyone has muted.
06
AI where it earns its place.
Roughly ten percent of what gets called "AI in crypto" is real work — flow classification, anomaly detection on chain, summarising counterparty comms. We do that ten percent and skip the rest.
02 — how we engage
We will not build an "AI trading bot" for retail crypto. We will not touch a token launch.
Typical scope is 3 to 10 weeks. We start with a week of reading your fills, your chains, and your incident log.
We work on the infra side of the desk: tooling, data, monitoring. The alphas stay yours.
03 — example shape
an engagement shape
Anonymised. Plausible. The shape of a real engagement.
A top-five crypto market-maker had inventory drift on three L2s and no clean view of why.
The indexer was lossy on reorgs and the dashboards aggregated stale balances.
We rebuilt the on-chain pipeline as a reorg-aware indexer with a 4-block confirmation window, and rewired the inventory view off the new schema.
Drift attribution dropped from "we think it is the bridge" to a per-pool breakdown updated every block.
The MM team killed two unprofitable pools the same week and redeployed the capital to a fourth chain.
The pipeline runs as a daily job. We were done in 6 weeks.
04 — contact
Tell us where the work is.
One sentence is enough. We reply within two business days. If we are a bad fit we will say so.