Backend Engineer (AI-Native) — Danger Infrastructure & Integrations at Coris

🧠 About Coris

Coris is the AI platform for threat and monetary operations. Platforms, marketplaces, cost suppliers, and banks – GoFundMe, Clio, Cherry, and a rising checklist – use Coris to onboard retailers sooner, monitor them repeatedly, and act on fraud with out including headcount. One buyer described us as “Cursor + Lovable for threat groups.”

We’re backed by Y Combinator, Lux Capital, and Pathlight, based by payments-domain specialists, and rising shortly on buyer pull. We’re a small group; this rent owns a big share of the backend.

👥 Who you will work with

Coris was began by Vinodh (CEO) and Shyam (CTO). We’ve every spent 20+ years in fraud and threat administration, working threat engineering, operations and product groups at Google, eBay, PayPal and WePay. You’ll work straight with each of us in Palo Alto.

🚀 Why this function issues

Every Coris determination runs on information we pull from cost processors and platforms: Stripe, Adyen, Fiserv, TSYS, Repay, and extra, over APIs, webhooks, and SFTP file feeds – tons of of hundreds of thousands of occasions a month, billions of rows in Postgres. On high of that sits a indicators and guidelines engine, an account graph, and the partner-facing API our prospects construct on.

Here’s the trustworthy model of the work. Files arrive out of order over SFTP. One processor sends dates in UTC, one other within the product owner’s native time, and neither says which. A processor’s decline code lives in a unique column than the spec says. A webhook ships malformed JSON at 2am. A backfill employee runs out of reminiscence on a 40B-row desk. A buyer’s threat lead says “our numbers look off” and desires a solution at this time.

We’re hiring the engineer who will personal all of that end-to-end, and who thinks it is the attention-grabbing half – as a result of each integration we make reliably is one a competitor hasn’t, and that is the moat. Getting the boring work proper each time is how enduring firms are constructed!

🥷 What you will do

• Own processor and platform integrations. Build and function ingestion from Stripe, Adyen, Fiserv, TSYS, Repay, and new processors as we signal them: APIs, webhooks, and batch file feeds. Make them appropriate, idempotent, observable, and PCI-safe.
• Own the accomplice API and customer-specific work. Extend the v2 API, construct per-customer configuration and indicators, and work straight with prospects’ threat and ops groups to show a imprecise report right into a repair, quick.
• Build the chance engine. Add indicators and guidelines to the DAG engine that scores retailers and transactions; design Postgres schemas, indexes, partitions, and aggregations that keep quick at billions of rows.
• Run it in prod. Celery employees, backfills, deadlocks, circuit breakers, latency, Sentry monitoring. You ship it, you personal it.
• Use AI in every single place it earns its place. Coding brokers for first drafts and refactors, AI evaluation on each PR, and LLMs contained in the product (dispute brokers, doc parsing, structured extraction). You’ll additionally assist set the group’s requirements for a way we use these instruments.
• ~10–20% product-facing TypeScript/Next.js when a function wants it.

📅 Your first 90 days

• Week 1: your first PR is merged and working in manufacturing.
• 30 days: you are choosing up substantial options and transport them.
• 60 days: you take on more durable issues and constructing actual depth in a number of elements of the system, like processor integrations, the chance engine or the accomplice API.
• 90 days: you are the go-to particular person for a major a part of the platform.

✅ You could also be a match if

Nobody matches each line under. If most of it sounds such as you and the work above is the type you wish to do, we would like to listen to from you.

• A number of years transport manufacturing Python backends (Django/Postgres or shut; 3+ years is a information, not a cutoff). Ideally you’ve got owned a system in manufacturing by way of progress: schema adjustments on giant tables, concurrency and locking bugs, backfills, on-call. You’re at dwelling in SQL and might learn a question plan, or wish to get superb at it.
• An AI-native method of working. You attain for coding brokers (Claude Code, Codex, Cursor or comparable) by default, and you have constructed one thing with LLMs that individuals use, at work or by yourself. If you are earlier on this curve however transferring quick, inform us what you’ve got constructed.
• Judgment about what AI produces. You have a really feel for which adjustments to let an agent run with (take a look at scaffolding, a well-specified refactor, a migration you’ll be able to diff) and which to learn line by line (something touching cash, card information or PII, a schema change on an enormous desk, locking and concurrency). You wish to perceive each line of a diff earlier than it ships, and also you deal with AI evaluation feedback as enter, not instruction.
• You just like the schlep. Integrations, escalations, backfills, file codecs from 1995. You see the sample within the mess and do not burn out on it, since you perceive why it is the product.
• Comfortable with prospects. You’d be comfy, or eager to get there, on a name with a buyer’s threat group, saying “give me an hour,” and coming again with the reply.
• Committed to the startup final result. You need fairness that issues and you’re employed like an proprietor: comfy with ambiguity, with lengthy weeks when it counts, and with being within the room 5 days every week.

🙏 Nice to have

• Payments, service provider threat, underwriting, or financial technology background (Stripe, Adyen, Plaid, Ramp, a PayFac or ISO).
• Prior early-stage startup expertise.
• Built a guidelines engine, sign pipeline, entity-linking / graph system, or function retailer.
• SFTP and batch-file integrations, fixed-width or legacy codecs, PCI-scoped information dealing with.
• Shipped LLM options in manufacturing – structured extraction, brokers with instruments, evals.

🤝 How we work

• Bias towards motion and measurable affect; we ship quick, measure, and iterate with actual buyer suggestions.
• Everyone owns their code in prod.
• AI-first growth: brokers write, people personal. Every PR will get an AI evaluation and a human one if mandatory.
• In-person, 5 days every week in Palo Alto. Like any high-growth startup, we go properly past the standard 40–50 hours every week.