Who gets hired and onto which teams
Mortgage servicing runs on mainframes older than the internet. Valon Labs decided to replace them with a single system of record — and that decision shapes every hire they make.
The company services more than $100 billion in mortgages on a platform built from scratch after starting as a services business to learn the domain, Valon's website reports. Founder Andrew Wang moved from mortgage-backed trading at Soros to building ValonOS, so the engineering organization sits inside a regulated financial operator, not a pure SaaS vendor. This guide covers who Valon recruits, what they earn, how the interview process works, where the work is performed, and which personal traits predict success.
ValonOS unifies three layers: a system of record for mortgage data, a workflow engine that orchestrates thousands of daily tasks, and AI agents that operate inside strict regulatory guardrails — explainability over autonomy, decision support over agentic freedom. The company's career page frames the filter: "exceptional individuals who care about getting it right, who want to think in decades rather than quarters."
Six years in, Valon reports 500,000 loans on platform, the Charles Rubenfeld interview states, and a 3× cost advantage over legacy servicers, Valon's data shows. Open roles on the Zero G Talent board show hiring weight toward platform hardening, AI agent reliability, and go-to-market muscle to onboard the next tier of servicers.
Compensation at a glance
Valon Labs' compensation data from that board shows a salary band running from $137,000 to $284,000, with a median of $235,000 across 18 salaried roles. The spread reflects a team weighted toward senior engineering and management, though the board also captures go-to-market and program-management functions. All figures below are drawn from live postings on that board; where a web source duplicates a number, the first-party board figure takes precedence.
| Role | Location | Salary band (USD/year) |
|---|---|---|
| Staff Software Engineer | New York | $255,000 – $300,000 |
| Engineering Manager, Platform - Cloud Infrastructure | Remote | $241,500 – $284,000 |
| Engineering Manager | New York | $241,500 – $284,000 |
| Strategic Account Executive | New York | $150,000 – $275,000 |
| Sales Engineer | New York | $150,000 – $275,000 |
| Technical Program Manager (Data) | New York | $185,000 – $250,000 |
Engineering roles sit at the top of the range. The Staff Software Engineer posting in New York carries the highest ceiling at $300,000, while the two Engineering Manager listings (one remote, one New York-based) share an identical $241,500–$284,000 band. That parity suggests Valon prices management scope consistently regardless of geography, at least for cloud-platform leadership.
The go-to-market roles show wider variance. Both Strategic Account Executive and Sales Engineer span $150,000–$275,000, a $125,000 spread. The Technical Program Manager (Data) slot lands at $185,000–$250,000.
Only one of the six sampled postings is explicitly remote: the Engineering Manager, Platform - Cloud Infrastructure role. The other five list New York as the work location. That aligns with the company's broader pattern of maintaining a New York hub while allowing select infrastructure-leadership positions to operate remotely.
That median across 18 roles indicates a concentration in the $200,000–$260,000 band. Few postings dip below $185,000, and none exceed $300,000 in base salary. Candidates should treat the published bands as base-compensation ranges; equity, bonuses, and benefits are not captured in the board data.
How the hiring process works
Valon's public careers page echoes that filter, seeking those who prioritize durable decisions over short-term wins. The company builds AI agents that must make auditable decisions inside one of finance's most regulated corners — mortgage servicing on a platform carrying $100B+ in loans. That environment rewards engineers who choose the harder, durable path over quick demos, who document reasoning so regulators can trace it, and who treat operational rigor as a feature. The company's policy restricting AI access for new hires underscores a culture that wants human judgment calibrated before automated tools enter the loop.
Specific interview stages, assessment formats, and candidate advice for Valon Labs are not documented in the available sources. The research contains a detailed interview account for a different company (Value Labs, an Indian IT services firm) that must not be conflated with Valon Labs.
Where the work happens
Valon Labs builds ValonOS, an AI-native operating system for mortgage servicing. It does not operate hardware labs, test cells, machine shops, or integration floors. The work happens on laptops and in cloud environments, distributed across a New York office and a remote-first engineering culture.
The company's job postings list five roles tied to New York and one Engineering Manager role explicitly marked Remote. That split reflects a hybrid model: a physical hub in New York for collaboration, recruiting, and in-person rituals, with the rest of the team distributed across time zones.
Valon's own site describes its platform as "software that scales with complexity instead of collapsing under it": a cloud-native stack orchestrating thousands of daily tasks through a customizable workflow engine. Engineers work in that stack: ValonOS, the unified platform that replaces fragmented legacy systems, providing a single source of truth for mortgage data, money movement, and compliance. The "test cell" is a staging environment; the "machine shop" is a CI/CD pipeline; the "integration floor" is a monorepo with feature flags and contract tests.
The company's blog notes the team is "just getting started: building AI agents that can make auditable decisions in one of the most regulated environments in finance is not a problem you can solve in a research lab." The lab, in this case, is the production environment itself, governed by the same auditability and governance controls the platform enforces for its servicer clients.
Carrington Mortgage Services, a residential mortgage servicer, lender, and insurance provider (NMLS #2600), partners with Valon Mortgage (NMLS #1907140) to run loans on ValonOS. That relationship means Valon engineers ship code that moves real money for real homeowners — 92%+ satisfaction versus a 70% industry average. The physical artifact of their work is not a rocket stage or a robotic arm; it is a ledger that stays correct, consistent, and reconciled at scale.
For a candidate, "where the work happens" translates to: a laptop, cloud infrastructure, and a team that treats mortgage servicing as a distributed-systems problem. The New York address is real; the remote option is real; the hardware lab is not.
Who thrives here
Valon's clearest signal about who lasts comes from a policy that looks backward to move forward. CEO Andrew Wang, a former Goldman Sachs analyst, instituted a rule last month: new hires in non‑engineering roles must learn their jobs without AI before they're allowed to use it. Engineers are exempt because every line of code passes peer review; finance, HR, and operations lack that guardrail. Wang told Business Insider the trigger was simple — employees were defaulting to the most expensive models for routine tasks, driving token spend toward $15–20 million annualized and, more critically, eroding the judgment needed to supervise the output. The policy cut projected AI spend to $4–5 million. It also forced a behavioral shift: junior staff started walking over to senior colleagues instead of prompting a model. That interaction (watching, asking, doing the basic work) is the training loop Wang wants to protect.
The company's own records bear this out. Experienced employees welcomed the restriction because they had been spending hours correcting low‑quality AI output from newer hires. A September 2025 BetterUp–Stanford survey of 1,150 desk workers found 40 percent had received AI‑generated work from a colleague in the prior month; each instance cost nearly two hours to check or rewrite. Valon's policy targets that hidden tax directly. The term "meat proxies" (coined by developer Niklas Gruhn for people who pass unchecked AI output downstream) captures the failure mode Wang is designing against.
Success at Valon clusters around three traits. First, comfort with regulated complexity. The platform services over $100 billion in mortgages on ValonOS, a system of record built from scratch to keep every transaction, balance, and audit event in a single source of truth. That demands engineers and product people who treat compliance as a design constraint, not a checklist. Second, willingness to earn expertise the slow way. Wang's blog post framed the policy as "weird for a company at the frontier of AI usage," but the logic is deliberate: if AI handles the entry‑level tasks that traditionally built judgment, senior roles get filled by people who never developed it. Third, mission alignment that survives friction. The company's 92 percent homeowner satisfaction (versus a 70 percent industry average), Valon's figures put, comes from a stack that unifies workflows, data, and money movement.
Wang has invited criticism: "If you have a much better idea here of how to make sure people learn, please tell me." No credible alternative has emerged. The trait that ultimately predicts longevity is the one the policy selects for — people who build understanding before they automate it.
Working in frontier tech? Zero G Talent tracks the openings: see every open Valon Labs role, browse frontier tech jobs, the companies hiring, and the people building the field.