The Platform Rewrite
PayU's Payments Intelligence suite consolidates fraud prevention, tokenization, and conversion optimization into a single API layer for Latin American merchants. The suite builds on an expanded Feedzai partnership across EMEA and LATAM announced in 2020. It targets regional gaps with AI analytics that model consumer behavior across Africa, Latin America, and Central and Eastern Europe; preventive high-risk transaction blocking before authorization; and a 3D Secure engine that dynamically routes authentication challenges based on issuer performance. PCI DSS Level 1 compliance and tokenization across web and mobile are included.
The architecture shifts toward continuous learning. Feedzai's engine, integrated since 2020, ingests transaction streams in real time and outputs dynamic risk signals. PayU's figures show the 3DS optimization alone has saved roughly one million transactions and $30 million for more than 50 European merchants. Oriflame, the beauty retailer, recorded a 20 percent lift in overall approval rates, a 21 percent year-over-year gain in buy-now-pay-later approvals, and an 8.8 percent card approval increase after deploying the 3DS routing.
A distinguishing module is "local payments with smart insights," a recognition that in Latin American markets, the payment method is a key risk vector. The platform maps approval and decline patterns across local rails, then surfaces actionable recommendations. Merchants gain a unified view across countries and payment methods, with benchmarking against industry trends to flag inefficiencies. The suite is currently recruiting design partners, signaling late-stage validation before broader rollout. For the region's fintechs and the engineers who build for them, the launch reframes fraud from a cost center into a data product, and that reframing is reshaping hiring plans.
Why ML Engineers Now Cost a Fortune
Toku, a Singapore-incorporated AI-powered customer experience platform, Toku reported 13 percent year-on-year revenue growth to $18.8 million in the first half of 2026. Its order book hit $29.3 million by June 2026, and Tier 1 customers (accounts above $500,000 annual revenue) more than doubled. Accounts with the AI suite deployed recorded a 26 percent uplift in monthly recurring revenue versus pre-deployment baselines, and gross margin on new bookings reached 89 percent. That trajectory requires ML talent to maintain transcription, summarization, sentiment analysis, and governed virtual agent capabilities across complex multi-market deployments.
Separately, a Latin America-focused recurring payments platform also named Toku secured $9.3 million in April 2024, according to a PR Newswire report, led by Gradient Ventures to fuel expansion in Mexico, Brazil, and Chile. Gradient Ventures partner Zachary Bratun-Glennon noted that "LatAm digital commerce grows rapidly, businesses and consumers face challenges in payment processing and communications."
Zero G Talent's live board data captures the salary pressure. Stripe (a global benchmark) posted 48 roles in the past seven days alone, including a Machine Learning Engineer position in South San Francisco banded at $212,000–$318,000. The board's overall salary band runs $49,000–$289,000 with a $235,000 median across 20 salaried roles. While those figures reflect U.S. locations, they set a reference ceiling for Latin American fintechs competing for senior ML engineers who can work remotely for global firms.
| Role | Location | Salary Band (USD/year) | Source |
|---|---|---|---|
| Machine Learning Engineer | South San Francisco, CA | $212,000–$318,000 | Zero G Talent board (Stripe) |
| Senior Software Engineer | South San Francisco, CA | $190,400–$285,600 | Zero G Talent board (Stripe) |
| Technical Program Manager, Infrastructure | South San Francisco, CA | $173,400–$260,200 | Zero G Talent board (Stripe) |
| Board median (all roles) | Global | $235,000 | Zero G Talent board (Stripe, 20 roles) |
Latin American fintechs face pressure from global companies hiring remotely at San Francisco bands. Mercado Pago, among others, is building internal AI teams. PayU's suite, by packaging capabilities into developer-ready APIs, has made the skill set more visible and more valuable.
The Arms Race: Horizontal vs. Vertical AI
Stripe moved first. In August 2026, the $159 billion payments giant announced a $7.5 billion acquisition of OpenRouter, an AI routing startup that sits between applications and large language models. The deal arrived weeks after Stripe and Advent International tabled a $53 billion joint bid for PayPal — a transaction that would hand Stripe hundreds of millions of PayPal accounts bolted onto merchant rails it already runs. Stripe processes $1.9 trillion in annual volume; the OpenRouter purchase signals a shift from facilitating payments to metering traffic between AI agents. As an A2Z Fintech interview noted: "The next decade of commerce is about agents transacting with agents, software buying compute, buying services, buying from other software. Whoever sits at the toll booth between the companies and the AI models meters the traffic and eventually moves the money. Stripe just paid $7 billion for a toll booth."
Stripe's AI push predates the OpenRouter deal. In May 2025 Stripe unveiled a foundational model trained on payments data and introduced stablecoin-powered accounts. By September 2025 it had co-developed the Agentic Commerce Protocol with OpenAI, enabling Instant Checkout inside ChatGPT. Its antifraud product Radar has operated since 2018, but the new model layer and the OpenRouter acquisition suggest a strategy of owning the intelligence layer, not just the risk layer.
Regional players are responding. dLocal, the Uruguay-founded, Nasdaq-listed processor, operates in the same Latin American markets. Mercado Libre, a major Stripe customer, has invested heavily in the region. The pattern is clear: global players buy or build horizontal AI layers, while regional players weaponize local data. PayU's AI-ready developer tools arrive into a market where serious competitors have shipped AI fraud upgrades, acquired AI startups, or embedded models into their core acquisition loops. The hiring surge reflects cumulative labor demand from an arms race already underway.
Can Regulation Keep Up?
PayU's AI fraud suite lands in a Latin American financial system rewiring itself around machine learning. The Saudi Central Bank's November 2026 Riyadh meetings, themed "Shaping Tomorrow: Harnessing Experience and Charting the Future of Innovation, Collaboration and Digital Transformation," signal where regulatory templates may emerge. The practical effect: any acquirer rolling out PayU's suite (or a competitor's) will need audit-ready model cards from day one, raising the bar for vendors and engineers.
Venture capital is following the signal. Toku's 1H2026 results, as reported by Toku, bore out the thesis: revenue rose 13 percent year-on-year to $18.8 million, Tier 1 customers more than doubled, accounts with the AI suite recorded the same 26 percent uplift in MRR against pre-deployment baselines, and gross margin on new bookings hit 89 percent, up from 56 percent a year earlier. Mercado Libre and dLocal self-fund AI as core IP. Early-stage startups like Toku raise from AI-specialist VCs (Gradient, F-Prime, Clocktower) to build vertical solutions that embed fraud scoring as a feature. In the middle sit regional acquirers and mid-market PSPs: they lack the balance sheet to build and the niche to specialize. PayU's bet is that this middle adopts its suite. The hiring data suggests the bet is working, but the regulatory timeline is a wildcard.
The Pipeline Is the Asset
The hiring spike triggered by PayU's AI fraud suite and competitive responses from Stripe, dLocal, and regional fintechs has exposed a structural gap: Latin America produces strong engineering graduates, but the specialized machine-learning talent required to productionize fraud models at scale remains scarce. Stripe's board data shows the intensity: 48 roles added in a single week, including machine learning engineers banded at $212,000–$318,000 and business systems architects at $274,000–$334,600. Those compensation levels, median $235,000 across 20 salaried roles, signal what it costs to attract experienced practitioners today. But they also reveal the ceiling of a strategy built purely on poaching.
The region's fintechs have grown faster than the local talent supply chain can feed them. Toku illustrates the operational reality: its 1H2026 results show revenue of $18.8 million, a 25 percent order-book increase to $29.3 million, and an 89 percent gross margin on new bookings, driven by an AI suite that lifted monthly recurring revenue roughly 26 percent post-deployment. Toku's response to the talent constraint has been "targeted recruitment and selective subcontracting," a phrase that describes the industry's current default, not a long-term solution.
The gap between academic coursework and production-grade ML engineering remains wide. The fintechs that move first to close it — through structured apprenticeship programs, co-designed curricula, and dedicated research partnerships — will convert the hiring surge into a durable advantage. Those that treat talent as a spot market will keep paying escalating premiums for a shrinking pool of senior engineers.
The regulatory environment adds another dimension. As compliance burdens on AI models grow, teams that can build explainable, auditable fraud systems become a regulatory asset, not just a product asset. That specialty, ML engineering fused with financial-regulation fluency, does not exist in the labor market today. It has to be grown.
PayU's platform rewrite began with a single API layer. The next decade will be written by the engineers who build the pipelines to feed it. The current hiring rush is the down payment. The pipeline is the asset.
Working in frontier tech? Zero G Talent tracks the openings: see every open Stripe role, browse frontier tech jobs, the companies hiring, and the people building the field.