Backend Engineer

Job Description

We’re looking for a passionate Backend Developer to join our growing engineering team. You'll be working on scalable backend systems, collaborating with cross-functional teams, and shipping real products used by thousands of users.

You build and extend the pricing and matching core — the product's IP.

Pricing engine. Implement the coupled simplex maker from a precise spec:

• The three probabilities as a single log-odds vector q = (q_KTL, q_TIE, q_GTL), with prices

as p = softmax(q) so they sum to 1 automatically and no buy-all / sell-all arbitrage exists.

• A flow nudge (δ = 0.05) that shifts an outcome's log-odds on filled flow, with the softmax

coupling automatically lowering the other two outcomes in proportion.

• A model/flow blend q_blend = w·q_model + (1 w)·q_flow− , where the weight w [0.25,∈

0.92] drops toward observed flow when a per-outcome imbalance crosses the toxicity

threshold (tox_thresh = 0.6).

• A dynamic half-spread that widens with toxicity (base 6.5¢, up to 16.25¢) and hard price

clamps (floor 3¢, ceiling 97¢), with a 10,000-share hard cap per trade.

• The workbook's four no-arbitrage checks wired as runtime assertions that halt the market

and page on-call when violated.

Matching tiers (the documented build order):

• Tier 1 — direct FIFO matching (same outcome, same YES/NO, opposite side): zero

maker risk, peer-to-peer.

• Tier 2 — intra-synthetic matching (YES_X ↔ NO_X economic equivalents): closes intra-

outcome flow book-to-book.

• Tier 3 — cross-outcome hedge matching, hedge-aware and L2-strict: pairs cross-

outcome orders only when the pairing strictly reduces the maker's L2 norm.

Maker-risk mechanisms that run alongside matching:

• Partial-fill throttle — binary-search the largest fill that keeps L2 at or below the exposure

cap; this is the system's non-negotiable safety net.

• Whale splitting (500-share chunks) — the single highest-leverage feature on cancel rate

and revenue; each chunk runs the full pipeline so maker depth builds between chunks.

• Maker auto-quotes — self-unwinding _pPost-tagged ladders ([100, 150, 200]) posted on

the unwinding side when |position| > 80.

• Mean-reversion / proactive unwinding with accelerated decay (scaling from a 7% base

toward a 25% cap as exposure grows) and inventory skew and a book-depth incentive

(rest/maker split that gets aggressive when a book is thin).

You'll measure everything the way the report does — cancel rate, U (residual maker

absorption), peak L2, and peak/1K — and reproduce the source exactly: Excel pricing-row

parity, the six shock scenarios, the 24 whale round-trips (the whale loses every config), and the

50×50 simulation metric envelope, all green in CI. A central, explicit unknown is adverse

selection: the simulations used random traders, and the live market is the first encounter with

price-responsive humans — laddered quotes can telegraph maker exposure, and the

documented safe fallback is to keep accelerated decay and revert to a single unwind quote.

Strong fit: quantitative / market-microstructure background, numerical-precision instincts,

comfort turning a mathematical spec into deterministic, test-covered code.

Requirements

• Solid backend engineering in TypeScript / Node.js (or strong adjacent experience and

the appetite to be fully productive in TS — the whole stack is one language, with shared

types across engine, API, and frontend).

• Comfort working from a written spec with test vectors and a habit of proving correctness

with tests rather than asserting it.

• Experience with PostgreSQL and event-driven architectures; an understanding of why

determinism, idempotency, and append-only logs matter here.

• A bias toward fail-safe design: when something is wrong, stop — never continue wrongly.

Nice to have:

• Prior work on an exchange, order book, trading, betting, or payments system.

• Quantitative / market-microstructure exposure, market-maker inventory-risk models, or

numerical optimization.

• Production WebSocket / streaming experience at scale, NATS or Kafka.

• Double-entry accounting or ledger-system experience.

• Familiarity with AWS (EKS, RDS), Redis, and Datadog/Sentry observability.

Stack:

TypeScript / Node.js · PostgreSQL (multi-AZ) · NATS JetStream · Redis · WebSockets · AWS

EKS / RDS · Terraform · Datadog · PagerDuty · Sentry

Benefits

• In addition to a market competitive compensation, we have a reward philosophy that expand beyond this.
• Fully remote
• Opportunity to work with a truly global team
• Flexible timings. You decide your work schedule