Agentic Execution Engineer

**ScribeUp** Agentic Execution Engineer; $200K–$250K + 0.10%–0.50% equity; United States (Remote)

**About Company** ScribeUp is a Series A embedded fintech building AI-powered bill-management infrastructure for financial institutions. Its product lives inside banking apps as a modern "bills tab". Recurring bills (rent, utilities, loans, insurance, subscriptions) are automatically detected and tracked, users get alerts before bills are due, and they can take one-click action natively in their bank's app: cancel, update the payment method, or save money on a bill. When ScribeUp goes live at a financial institution, around 60% of end users immediately find a bill they didn't realize they were paying.

The platform rests on two core technical competencies. The first is intelligence built on full banking context, which over-the-top apps can't match. The second is agentic execution that can act on essentially any biller. ScribeUp has signed 60+ financial institutions in about 18 months, from neobanks and fintechs such as DailyPay and NerdWallet to banks and credit unions across the US. It is pre-integrated with the leading digital-banking providers (Q2, Alkami, Jack Henry/Banno, Lumin, Candescent) and recently raised a $10M Series A to scale Product, Engineering and Go-to-Market.

The team is high-caliber, low-ego and flat, with minimal micromanagement: talented people get outsized responsibility and clear ownership lanes. ScribeUp starts remote, with hybrid (~3 days/week in office) preferred but not required and a New York office planned depending on roles landing there in broader Q1 2027 hiring, and runs regular on-sites and virtual team events.

**About the Role** This engineer joins ScribeUp's agentic execution team, which builds Iris, the engine that carries out bill actions for users directly inside their banking app, and reports to the Director of Engineering. Here's how it works: a user taps "cancel" on a bill and signs in to the merchant through a headful browser session. Iris then takes over the authenticated session, navigates the merchant's flow, executes the action and verifies it actually completed. On merchants it has never seen, Iris runs agentically. Successful runs are turned into faster, cheaper deterministic instructions, and Iris falls back to the agent when a merchant's flow changes. Iris is live in production inside regulated financial-institution experiences, and the mandate is to extend it to any action on any biller. This is not a research role or a thin wrapper on someone else's agent framework.

**Responsibilities** - Own the systems that execute live merchant actions (cancellations, card switching, plan changes) end to end, and raise completion rates and accuracy. - Own the loop between agentic and deterministic execution, including the eval gates that decide when an agentic run can be trusted enough to become a deterministic instruction set. - Grow merchant coverage from a few hundred directly connected billers toward any biller. Build merchant classification and coverage tiering against anti-bot defenses, and pick the execution stack, browser configuration and network strategy for each tier. - Own the headful browser and session-handoff architecture: session management, proxy/network egress, and resilience against detection. - Own the verification layer, plus the logging, run capture and observability behind it, so Iris never reports an action it didn't complete. - Expand Iris into new actions (bill savings, pausing, plan changes) in partnership with Product.

**Required Skills** A. Hands-on production experience with browser automation tools (e.g. Playwright, Puppeteer, Selenium or headful/stealth browsers) at scale. B. Has built agentic products with LLMs and agent frameworks where the agent is the product, not just used AI coding tools. C. Strong software engineering fundamentals with deep proficiency in Python or a comparable language. D. Ships fast while keeping the QA discipline a regulated financial environment requires.

**Bonus Skills** - Adversarial/anti-bot environments: stealth browser tooling, proxy infrastructure or bot-detection evasion at scale. - Session-handoff or token-based authentication architectures. - Turning LLM outputs into deterministic, auditable artifacts (instruction sets, evals, verification loops). - Fintech, payments or other regulated-environment experience (SOC 2, PCI). - Distributed systems/microservices on AWS or GCP. - Early-stage startup experience; experience hiring and growing an engineering team.

**Logistical Info** - Location: US-based, full-time. Starts remote; hybrid (~3 days/week in office) is preferred but not required, with a New York office planned depending on roles landing there in broader Q1 2027 hiring. - Compensation: $200K–$250K base + 0.10%–0.50% equity (fully diluted). - **Other:** Reports to the Director of Engineering; team on-sites for remote and in-office staff.

**Ideal Background** - 🔴 Please don't outreach on our behalf to Pinwheel (most important), Alkami, Candescent, Banno (part of Jack Henry), Q2, StrategyCorps, or DailyPay (a ScribeUp client, conflict of interest). - The main qualifier is browser automation. The client was explicit that the candidate can come from any industry or product, as long as they are a strong software engineer with hands-on browser-automation experience building agentic products. - Strong fits: engineers from RPA, web-automation/scraping infrastructure, AI browser-agent companies, or teams on either side of bot detection. - Up to date on the latest agent tooling. The client would rather have someone current than someone relying on older approaches, since the space changes monthly.

**Green Flags** - Has owned completion-rate, accuracy or verification metrics for an automated system in production. - High personal ownership; comfortable operating where no playbook exists yet. - Resonates with the mission of giving consumers control over their recurring bills.

**Red Flags** - 🔴 Please don't outreach on our behalf to Pinwheel (most important), Alkami, Candescent, Banno (part of Jack Henry), Q2, StrategyCorps, or DailyPay (a ScribeUp client, conflict of interest). - Currently works at a credit union (e.g. Navy Federal, PenFed, BECU, Alliant, or any local credit union) — ScribeUp's prospective clients; do not contact. - Currently works at a bank (e.g. Chase, Bank of America, Wells Fargo, Capital One, or any regional or community bank) — ScribeUp's prospective clients; do not contact. - Currently works at a consumer-facing fintech (e.g. Chime, SoFi, Cash App, Venmo, or similar apps people use to bank, pay, or save) — ScribeUp's prospective clients; do not contact. If unsure whether a company counts, ask before reaching out. - No hands-on browser-automation experience, even with strong general AI/ML or backend credentials. - Research-only background, or has only built thin wrappers on third-party agent frameworks. - AI experience limited to using AI coding assistants rather than shipping agent-driven products.