$200K–$300K salaries advertised for CollectWise’s four NYC AI Agent Engineer roles

A Five-Person Team Rewrites the Playbook

A five-person team just took a $35 billion industry's playbook, Y Combinator reported, and rewrote it with voice agents that outperform human collectors two to one, Y Combinator's data shows. Now the company behind that shift is hiring for the infrastructure that will carry it to ten times its current revenue.

CollectWise, a Y Combinator F24 company founded in 2024, has posted four open roles across engineering and commercial teams. The listings appear on the company's Y Combinator jobs page, on The Jobs Map, and on tryjeremy.com, all dated August 2026. Founder and chief executive Sean O'Brien says the team scaled to a $2 million annualized run rate in a few months, Y Combinator found, and now targets $10 million within the year, according to Y Combinator.

The four positions span the stack powering CollectWise's autonomous recovery platform. An AI Agent Engineer role carries a $200,000–$300,000 base plus 0.25–1% equity. A Senior Forward Deployed Engineer lists $240,000–$300,000. A Founding Account Executive shows the same $200,000–$300,000 band. A separate Forward Deployed Engineer posting on tryjeremy.com indicates $160,000–$250,000. All four are full-time, New York City-based roles.

CollectWise's site claims its AI agents deliver five times average ROI, CollectWise's site reported, 80% lower costs, CollectWise's figures show, and 100% account coverage while maintaining compliance across federal and state regulations. The Jobs Map describes the platform as letting creditors and agencies double recovery rates while cutting collection costs by three-quarters, The Jobs Map found. O'Brien frames the market as a $35 billion opportunity in the United States alone.

Y Combinator's directory lists CollectWise with twelve people and active status. The Jobs Map shows eight. The discrepancy reflects the hiring pace now underway. The AI Agent Engineer posting emphasizes voice AI infrastructure built on LiveKit, prompt engineering for compliance-safe conversation flows, and evaluation frameworks tying technical metrics directly to liquidation rates and containment. The forward-deployed roles focus on migrating clients from legacy voice infrastructure and translating business goals into agent behavior.

The next section breaks down what the most heavily posted role actually requires and why the skill set is narrower than the titles suggest.

One Role, Four Slots

The research captures a single, heavily detailed role, AI Agent Engineer, posted across Y Combinator's Work at a Startup board, OnlyCoders, Clera, and TryJeremy between February and August 2026. CollectWise lists four openings, but every public posting describes the same title and virtually identical responsibilities. Whether the four slots represent multiple hires for this one profile or three additional roles that haven't yet surfaced, the AI Agent Engineer specification is the only concrete footprint available.

The August 2026 Y Combinator posting, the freshest source, frames the role as building "the infrastructure, prompting systems, and core architecture behind our voice AI agents" at the intersection of "AI infrastructure, prompt engineering, conversational design, and product development." The engineer owns the end-to-end prompting strategy: conversation flows, tone, compliance-safe phrasing, guardrails, and edge-case handling. They also design and own the real-time voice agent architecture — orchestration, state management, latency optimization, and reliability — built on LiveKit. Testing and evaluation frameworks fall to them too: A/B testing, regression testing, conversation QA, benchmarking, and KPI reporting tied to containment, conversion, liquidation rates, and customer satisfaction.

Technical requirements are specific. The August posting asks for two-plus years in voice AI, conversational AI, prompt engineering, speech systems, or related agentic workflows, plus three-plus years in platform, infrastructure, or backend engineering. Strong Node.js, AWS, and SQL experience is required, along with production startup experience. Preferred signals include LiveKit or similar voice orchestration platforms, SIP trunking and telephony infrastructure, real-time audio pipelines, call routing, carrier integrations, and speech infrastructure in production. The stack across postings includes AWS, TypeScript, React, PostgreSQL, Node.js, Firebase, SQL, LLM prompting, A/B testing, AWS Lambda, OpenAI, conversational AI, and TTS.

The February 2026 OnlyCoders posting, six months older, describes a narrower slice: designing, optimizing, and productionizing the prompting and conversation logic behind voice AI agents while supporting the technical systems that power customer deployments. It emphasizes prompt development, testing, and iteration loops that improve identity verification, payment conversion, dispute handling, and containment rates, with close collaboration with the founder and customers. Requirements overlap heavily: two-plus years in voice AI or prompt engineering, strong analytical and business mindset, modern backend frameworks (Node.js), AWS, SQL, production environment experience, and excellent communication. That posting adds "Experience with GPT-5 or other LLMs", a forward-looking line absent from the August version.

Compensation differs between the two postings. The August YC listing quotes $200,000–$300,000 base with 0.25–1% equity. The February OnlyCoders listing quotes $150,000–$200,000 base with the same equity band. The gap likely reflects a re-leveling as the company scaled from a three-person team at $1 million ARR to a five-person team at $2 million ARR targeting $10 million ARR.

What the research does not show: distinct postings for an ML specialist, a legal-integration expert, or a product manager. The AI Agent Engineer role itself absorbs product collaboration, compliance phrasing, and evaluation framework design: functions that at larger firms might split across multiple hires. If CollectWise's four openings include specialized roles, they have not yet appeared on the primary boards the research covers. The evidence points to a single, high-leverage engineering profile blending voice infrastructure, prompt engineering, telephony integration, and business-outcome ownership, exactly the hybrid that autonomous debt-collection platforms need to ship compliant, revenue-generating agents at scale.

The Talent Market Bends Toward Compliance

The debt collection software market is projected to expand by $2.3 billion by 2028, and CollectWise's hiring push arrives as the sector's AI adoption curve steepens. A June 2024 Panorama Legal survey found nearly 40% of industry respondents already identify AI as a significant area of opportunity. Meanwhile, 78% of consumers now block calls from third-party agencies, forcing a wholesale shift toward digital, omnichannel engagement that only autonomous agents can scale. McKinsey estimates that agencies deploying generative AI can cut operational expenses by roughly 40% and lift recoveries by about 10%, with customer satisfaction scores climbing up to 30%. Those economics turn every AI engineer who can build compliant, explainable recovery agents into a high-leverage hire.

Visa's experience illustrates the broader fintech dynamic. Jose Lobez, the company's vice president of global AI and data innovation, said at the 2024 MIT FinTech conference that rolling out AI tools has increased hiring rather than reducing headcount, because the technology accelerates innovation instead of merely cutting costs. Visa backs that posture with roughly 2,000 fintech partnerships and a $100 million generative AI fund. CollectWise competes for the same talent pool — AI engineers fluent in NLP, predictive modeling, and regulatory guardrails — while also needing domain specialists who understand the explainability requirements regulators now demand. MIT's Tewary observed that customer-facing AI in financial services faces "an absoluteness" from regulators: you must show why a decision was made, trace the data lineage, and validate outputs against hallucination risk.

Salary benchmarks reflect that scarcity. Benchmark data shows Stripe's median posted salary at $235,000 (band $120K–$286K) across 19 salaried roles added in the past week, while ASML's median sits at $164,000 (band $31K–$258K) across 31 roles. Neither company operates in debt recovery, but both set the fintech/AI compensation floor that CollectWise must meet or beat. Indeed listed 44 AI debt collection openings as of the latest scrape — roles titled Senior Data Scientist, Data Collector, and Debt Collector — signaling that demand has moved beyond pilot projects into production hiring. Annual software licensing costs of $480–$1,200 per user, cited by Investors Hangout, further pressure firms to automate: the ROI on a single AI engineer who can replace or augment multiple collector seats compounds quickly.

The competitive ripple extends to incumbents. AgreeYa Solutions and Experian already market AI-driven collection suites, and large enterprises buy for economies of scale. Startups with API-first architectures, the profile CollectWise fits, gain an edge because established players need integration speed. Tewary noted that big finance "wants to work with startups with an architecture that's API-driven."

The Compliance-Engineering Hybrid Emerges

The AI Agent Engineer posting makes the new skill set explicit: prompt versioning, real-time voice orchestration with guardrails, evaluation against liquidation metrics. These requirements are portable to healthcare scheduling, insurance claims, and any regulated outbound voice use case.

The evaluation mandate shifts focus from model training to conversation-level metrics. The role asks for someone who can translate compliance requirements into deterministic guardrails that an LLM prompt can enforce at inference time. Candidates with backgrounds in regtech or bank compliance engineering are reframing their work as prompt-level policy compilation.

CollectWise's four postings represent more than headcount growth. They map the architecture of a new engineering discipline: voice infrastructure that treats compliance as a runtime constraint, not a checklist. The five-person team that rewrote the $35 billion playbook now needs the engineers who can keep rewriting it, at ten times the scale, under the same absoluteness regulators demand.

Working in frontier tech? Zero G Talent tracks the openings: see every open ASML role, browse frontier tech jobs, openings at Stripe, and the people building the field.