New York City's financial-domain AI evaluators are earning $220K–$285K in total compensation — a 35–74% premium above the $163K generalist evaluator median tracked in ENTRA's Salary Survey Q2 2026 (n=2,140). The premium is not a temporary artifact of demand outpacing supply; it is the output of a structural condition that no other US city can replicate: Manhattan and its surrounding metro contain more than 350,000 licensed financial professionals — FINRA registrants, CFA charterholders, Series 7/63 holders — within a 30-mile radius, supplying the raw material for domain-expert annotation at a density that San Francisco's generalist tech talent pool cannot match. As SEC AI model disclosure requirements take effect in Q1 2027, pushing banks to document RLHF training datasets for regulatory audit trails, that structural advantage is converting from a market edge into a compliance requirement.
At a glance
- NYC holds 18% of all US financial-domain AI evaluator postings in Q3 2026, second only to San Francisco at 23% — ENTRA Job Signal Index Q3 2026 (47,000+ postings tracked)
- Financial-domain evaluators at NYC banks command $220K–$285K TC, against a $163K generalist median — a 35–74% premium — ENTRA Salary Survey Q2 2026 (n=2,140)
- NYC AI trainer job postings grew 156% YoY in Q3 2026 — ENTRA Job Signal Index Q3 2026
- Bloomberg LP fields ~340 RLHF and evaluation specialists; JPMorgan AI Research at Columbus Circle employs 280+ AI training data specialists
What Happened: Three Institutions Built the Infrastructure
The East Coast AI evaluation economy was not assembled by a single employer decision. It consolidated around three institutional builds that each scaled past 100 roles in 2025 and accelerated further in 2026.
Bloomberg LP now runs the largest RLHF and evaluation operation of any non-lab financial institution in the United States. As of Q3 2026, approximately 340 specialists work inside Bloomberg's AI team on tasks that span preference labeling, output quality review, and model behavior evaluation for Bloomberg's financial intelligence products. Salary bands for that cohort run $185K–$265K base — a floor well above the generalist evaluator market, reflecting Bloomberg's requirement that evaluators carry working knowledge of fixed income, equities data, or derivatives market structure to perform meaningful evaluation of the company's AI-assisted terminal features. Bloomberg does not publish headcount breakdowns by function, but three former Bloomberg ML platform employees, speaking with ENTRA separately in Q3 2026, confirmed the scale of the operation and the qualification bar: candidates without CFA, Series 7, or equivalent demonstrated financial market expertise are screened out early in the process.
JPMorgan AI Research, headquartered at the firm's Columbus Circle campus in Manhattan, has built a 280+ person team of AI training data specialists. Total compensation for those roles — base salary plus the deferred cash and equity components that JPMorgan structures into technology compensation — runs $220K–$285K. The Columbus Circle location matters for recruiting: the campus places the team within walking distance of the firm's asset management, securities, and investment banking operations, enabling the knowledge transfer between practitioner domain experts and AI trainers that produces the high-specificity evaluation data JPMorgan's internal models require. JPMorgan's AI research output has focused heavily on document understanding, financial reasoning, and risk model audit — categories where domain-specific human evaluation is the bottleneck, not compute.
Goldman Sachs AI Platform has filled 120+ evaluation roles across its AI governance and model development functions, at compensation bands of $210K–$270K (per ENTRA Job Signal Index Q3 2026 LinkedIn headcount monitoring and sources familiar with Goldman Sachs's evaluation compensation structure). Goldman's AI evaluation work spans both generative model quality review for client-facing tools and more specialized red-team evaluation for trading infrastructure applications. The firm's decision to build in-house rather than outsource evaluation to platforms like Scale AI reflects a judgment shared across Wall Street: financial AI training data is too sensitive and too institution-specific to route through third-party annotation pipelines.
Taken together, these three institutions account for roughly 740 structured AI evaluation roles in the NYC metro — all paying above $185K base, none accessible through gig-economy annotation platforms, and all requiring domain expertise that New York's financial professional concentration makes easier to source than anywhere else.
Why It Matters: The Domain-Expert Moat
The generalist AI evaluator market is a global, increasingly commoditized labor pool. Annotators without specialized expertise compete on throughput and accuracy at rates that Levels.fyi Q3 2026 data puts at a $120K–$163K median for US-based workers — compressed from a Q1 2025 median of $134K–$178K as supply has grown faster than demand across the general annotation market.
The financial-domain evaluator market is structurally different. The job is not to assess whether a model output is coherent or grammatically correct. It is to assess whether a model's analysis of a corporate debt covenant is legally accurate, whether an AI-generated risk summary correctly characterizes the tail exposure in a derivatives book, or whether a model trained to assist compliance officers has encoded the right interpretation of Reg T margin requirements. Those tasks require evaluators who have spent years inside the financial system — reading the same documents, sitting in the same risk committees, and internalizing the same regulatory frameworks the model is expected to reason about.
New York's structural advantage is that it has produced those professionals at scale for decades. The 350,000+ licensed financial practitioners within a 30-mile radius represent not just a recruiting pool but an annotation quality guarantee: evaluators who flag a model's reasoning error do so because they have personally navigated the same reasoning problem in a live context, not because they read about it in training documentation.
The SEC regulatory driver reinforces the structural case. SEC AI model disclosure requirements, effective Q1 2027, require financial institutions deploying AI in client-facing contexts to document their model training processes — including the human feedback datasets used in RLHF pipelines. That documentation requirement creates an audit-trail demand that gig-economy annotation, with its informal worker relationships and limited provenance tracking, cannot satisfy. Structured employment of AI evaluators — the model JPMorgan, Bloomberg, and Goldman have all adopted — produces the documentation artifacts that SEC compliance will require. Banks building in-house annotation teams in Q3 and Q4 2026 are not just responding to model quality demands; they are pre-empting a compliance requirement that becomes mandatory in 90 days.
What's Next: Three Things to Watch
1. The mid-market bank build. JPMorgan, Goldman, and Bloomberg are first movers. Morgan Stanley, Citi, and Bank of America have smaller but growing AI evaluation teams — ENTRA Job Signal Index tracked a combined 87 new financial-domain AI evaluator postings from those three firms in Q3 2026, up from 31 in Q3 2025. The 156% YoY growth in NYC AI trainer postings overall is not yet saturated. The next wave of institutional hiring is the second tier of Wall Street firms reaching the same conclusion about in-house evaluation that the bulge bracket reached eighteen months earlier.
2. Scale AI's NYC financial vertical. Scale AI operates a financial services vertical out of its New York office that recruits domain-expert contractors for training data projects at banks and asset managers. The rates Scale AI offers for financial domain evaluation — not published, but estimated by three sources familiar with the market at $95–$140/hr for senior practitioners — are below full-time JPMorgan or Bloomberg compensation but above the platform's generalist rates. As the structured employment model becomes the compliance default, Scale AI's contractor arrangement may face pressure from the same SEC documentation requirements pushing banks toward in-house hiring. How Scale AI adapts its financial vertical to the audit-trail requirement is the firm's most consequential 2027 product decision.
3. The talent supply constraint. The 350,000 licensed financial professionals in the NYC metro are not uniformly available for AI evaluation work. Most are employed full-time in roles that preclude significant annotation side work. The actionable pool — mid-career finance professionals with sufficient domain depth and flexibility to join AI teams, or recently retired practitioners available on a structured basis — is smaller. As the institutional build continues, competition for that population will push compensation above the current $220K–$285K band. ENTRA Salary Survey data already shows the upper end of the Goldman and JPMorgan bands expanding: roles requiring deep derivatives or structured credit expertise are clearing at $295K–$315K TC in Q3 2026, above the published band ceiling.
New York did not plan to become the East Coast anchor of the AI evaluation economy. It became one because it already was the anchor of the financial expertise economy — and domain expertise is what the AI trainer economy values most. The premium is 74% above the generalist median, and the SEC clock is now running.
Global Career PlatformFind AI talent. Find your next role.
Booking is hotels. · Airbnb is apartments. · ENTRA is global careers.