The rhythm of work
Cresta's real-time AI platform sits inside the contact centers of United Airlines, Alaska Airlines, Cox Communications, and Intuit, listening to calls, coaching agents mid-conversation, and automating repeatable slices. The company operates remote-first across U.S. time zones, a structure visible in its own job listings: a Staff Machine Learning Engineer role lists "United States (Remote)" alongside hybrid options in New York and San Francisco, and the same remote designation appears on senior sales and marketing leadership posts.
Product development spans the core inference stack powering real-time guidance and automation, the platform layer handling data ingestion and tenant isolation, and the applied-AI team tuning models for each vertical's lexicon and compliance regime. A Staff ML Engineer posting at $230k–$300k signals the seniority Cresta expects on the modeling side.
Operations tempo is shaped by the enterprise sales cycle. The board lists three director-level sales roles (Enterprise West, Strategic West, Strategic East) each banded at $320k–$380k, indicating that landing and expanding Fortune 500 accounts drives quarterly planning. Engineering sprints frequently align to proof-of-value milestones: a custom integration for a pilot, a latency benchmark for a security review, a data-residency configuration for a regulated vertical. That customer-facing pressure creates a feedback loop where field engineers and product managers rotate into deal support, then carry the sharp edges back into the roadmap.
What's less documented in public sources is the internal ritual layer: how incident retrospectives run, whether design reviews are mandatory or advisory, how much autonomy a two-person pod has to refactor a data pipeline. The remote-first posture and the enterprise-grade SLAs suggest a bias toward written process and explicit ownership, but the specifics live in internal workspaces not visible externally. What the available signals do show is a company that has chosen velocity and customer proximity over organizational comfort, a trade-off that shows up in the compensation bands, the remote distribution, and the seniority of the roles they're still hiring for.
Values that show up in the product
Cresta's mission statement reads like a product spec: unlock the full potential of every customer conversation with human-centric agentic AI that delights customers, empowers humans, and turns strategic insights into business outcomes. That phrasing ("human-centric agentic AI") appears across the company's public pages and LinkedIn posts, and it functions as more than marketing copy. It sets a boundary condition for product decisions. When the team launched Opera, a no-code AI orchestration engine, they built it so non-technical operations leaders could configure workflows without engineering tickets. When they shipped AI Agent, they positioned it as automation that hands off to humans at the right moment, not a replacement loop. The Training Simulator, released in 2026, uses AI agents to simulate customers grounded in real conversation data so human agents can practice — augmentation, not substitution.
The principle "quick wins fund bigger transformation" shows up in how Cresta guides deployments. A July 2026 LinkedIn post described a customer that "focused on a use case that was simple to implement, delivered immediate value, and freed agents from repetitive work" rather than trying to solve every challenge at once. The same post framed the mindset bluntly: "You can throw a rock and hit a pain point worth automating. The question isn't where to start, it's whether you'll start at all." That bias toward starting small and compounding value maps to the company's own history: the first production deployment was a transformer-based Agent Assist at Intuit, a single digital channel use case, before expanding to voice, then to full conversation intelligence, then to autonomous AI agents.
"Every customer conversation contains signals that can shape your business" drives the Insights product line. The July 2026 launch of next-generation Cresta Insights emphasized detecting emerging issues in minutes, including "unknown unknowns," so teams can act faster. The upgraded AI Analyst, Topic Discovery, and Real-Time Trends capabilities are built to surface patterns without manual tagging, a direct translation of the value that conversations are data assets, not just support tickets. Customers including Oportun and Snap Finance cite QM automation gains and workload reductions of 50 percent or more, outcomes that only materialize if the platform actually catches signals humans miss.
Curiosity and courage to see what's next appears in the R&D trail. Cresta holds multiple conversational AI patents. It expanded from digital chat to Voice Agent Assist, a technically harder problem requiring real-time speech-to-text, latency management, and turn-taking logic. The company recruited Ping Wu, co-founder of Google's Contact Center AI, as CEO, a signal that the board wanted someone who had built at Google scale and understood the gap between demo and production. Doug Leone, formerly of Sequoia, was elevated to Chairman; Carl Eschenbach, former CEO of Workday, joined the board. These are not advisory appointments; they are operators who have scaled enterprise software businesses.
The tension between high velocity and principle-driven restraint shows up in the hiring data. The board's salary band runs $62k–$300k with a $205k median across 41 salaried roles, and recent postings include a Staff Machine Learning Engineer at $230k–$300k and a Head of Global Talent Acquisition at $220k–$300k. Those bands suggest a company paying for senior judgment, not just headcount growth. The remote-first model with 10-plus team hubs across APAC, EMEA, and the Americas means the values have to survive async communication and time-zone handoffs. "Combining the best of AI and human intelligence" is easy to say; doing it across a 600-person distributed team while shipping AI Agent, Insights, Simulator, and a major sports partnership (New York Mets) in the same year is the test.
Inside the interview loop
Cresta runs a four-round loop for virtually every role, but the content of those rounds shifts sharply by function, a signal that the company hires for craft depth first, then layers on culture fit at the end. OnJob.io's breakdown of current openings shows the pattern: Data Analyst candidates face SQL, Excel, a business case, and a visualization round; Sales and Business Development applicants move through a screening, a live role-play, a behavioural deep-dive on past quota attainment, and a manager conversation about CRM discipline and rejection handling; Software Engineers take a timed HackerRank assessment, one or two live DSA sessions, a system-design review (low-level for early-career, scalable-service for mid/senior), and a hiring-manager round focused on ownership and team fit; Backend Developers swap the initial screen for a language-and-runtime deep-dive before database/API design and system design; DevOps Engineers start with Linux internals and scripting, then CI/CD pipelines, cloud-and-container fluency, and a production-incident troubleshooting scenario. Glassdoor lists 53 interview reviews and 46 questions as of July 2026, and TeamBlind threads confirm the multi-stage cadence, though exact round counts vary by seniority.
What surviving each stage reveals is consistent across tracks. The first technical round (whether SQL joins, a coding assessment, or a Linux permissions drill) filters for baseline fluency and the ability to execute under time pressure. Candidates who advance tend to treat the problem as a proxy for the kind of constrained, measurable work Cresta ships: contact-center metrics such as call containment, agent efficiency, attrition, and revenue per conversation are all quantified, and the engineering team builds against those same hard numbers. The second round (Excel modelling, live DSA, language runtime internals, or CI/CD pipeline design) tests whether a candidate can reason through the layer beneath the abstraction. Ping Wu has said the company self-selects for people "excited about solving hard technical problems… in the context of can you driving towards customer value" rather than "we have a hammer and looking for nails." The third round (a business case, system design, or database/API architecture) is where product judgement shows up. Candidates who frame trade-offs in terms of agent productivity, customer satisfaction, or ROI tend to resonate; those who optimize for technical elegance without a clear line to contact-center impact often stall.
The final hiring-manager round is explicitly cultural. Wu has described the operating principle "lean" — literally crossing out "stay" from "stay lean" — and the conversation probes whether a candidate has operated in resource-constrained, high-accountability environments. Questions centre on past ownership: what you shipped, what you cut, how you measured success, and how you handled failure. The board's live postings (41 salaried roles with a $62k–$300k band (median $205k)) include senior titles such as Staff Machine Learning Engineer and Head of Global Talent Acquisition, suggesting the loop scales in rigour with scope. A candidate who clears all four rounds has demonstrated not just technical competence but a bias toward measurable customer outcomes, comfort with lean resourcing, and the communication clarity to sell a design decision to a cross-functional room. That profile matches the self-selection Wu described: people who want hard problems that pay off in the real world, not hard problems for their own sake.
Compensation: the 75th-percentile strategy
Cresta's compensation structure reads like a company that knows exactly which talent markets it's fighting in, and prices accordingly. The salary band stretches from $62,000 to $300,000 with a median of $205,000 across 41 salaried roles. That spread tells you everything about where the leverage sits: the top of the band belongs to revenue-critical sales leadership and specialized AI engineering, while the floor reflects earlier-career or support functions that don't appear in the recent postings.
| Role | Salary Band (USD) |
|---|---|
| Enterprise Sales Director West | $330,000–$380,000 |
| Strategic Sales Directors (East/West) | $320,000–$360,000 |
| Head of Demand Generation & Growth Marketing | $280,000–$330,000 |
| Staff Machine Learning Engineer | $230,000–$300,000 |
| Head of Global Talent Acquisition | $220,000–$300,000 |
Zero G Talent's board data shows the most recent listings cluster at the high end. Director-level sales roles sit at the ceiling. These are quota-carrying, enterprise-cycle owners selling into Fortune 100 contact centers (Alaska Airlines, Cox Communications, Intuit). The pay reflects deal complexity: multi-continent deployments, on-premise telephony integration, nine-figure ARR accounts. A Head of Demand Generation & Growth Marketing ranges $280,000–$330,000 across New York hybrid, San Francisco hybrid, and fully remote options — notably, no geographic discount for the remote track.
Zero G Talent's figures put engineering compensation signaling the same market awareness. A Staff Machine Learning Engineer lists at $230,000–$300,000 remote. That band aligns with late-stage AI companies competing for researchers who can ship transformer-based real-time inference at contact-center scale, Cresta's founding technical moat since the Stanford AI Lab days. The Head of Global Talent Acquisition at $220,000–$300,000 remote suggests the company treats recruiting as a strategic function, not overhead, consistent with a 600-person organization still scaling toward $100M ARR and beyond.
What's absent from the board data is as telling as what's present. No equity figures appear in the postings (typical for private-company listings), but the $280M raised across Series A through D (Greylock, a16z, Sequoia, Tiger Global) and a $100M ARR milestone imply meaningful option pools with secondary liquidity potential. Benefits details don't surface in the board feed either, though the remote-first posture (nearly every senior role offers a United States (Remote) option alongside hybrid hubs in those cities) functions as a de facto benefit: location arbitrage without salary penalty.
The philosophy reads clearly: pay at the 75th–90th percentile for the specific talent slice you need, keep geographic parity for remote workers, and weight variable compensation toward roles where revenue attribution is direct. Sales carries the highest ceiling because the product (a unified platform for human and AI agents across voice, chat, and omnichannel) demands consultative, multi-stakeholder selling into complex legacy environments. Engineering pays a premium for ML depth over full-stack breadth. And the median at $205,000 across all roles suggests a workforce skewed senior, which matches a company that has done 300+ contact center deployments across three continents and needs people who can operate without hand-holding.
If you're evaluating an offer, the band gives you a framework: ask where the role sits relative to the $205,000 median, whether equity is expressed as a percentage of the current preferred valuation, and how refresh cadence works post-Series D. The numbers are public. The negotiation starts from there.
Who stays, who leaves
Cresta's trajectory — from Stanford AI Lab spinout to $100M ARR serving United Airlines, Alaska Airlines, Cox, CVS, and the New York Mets — selects for a specific profile. The company's own language gives the clearest signal: "quick wins fund bigger transformation," that same principle, and the blunt challenge of that same challenge. That bias for motion, shipped product, and measurable customer impact defines who stays and who leaves.
People who thrive share three traits. First, they treat ambiguity as raw material, not a blocker. Cresta has launched Agent Assist, Voice Agent Assist, Opera (a no-code orchestration engine), AI Agent, a Training Simulator, and Real-Time Trends in roughly seven years, each requiring different technical stacks, buyer personas, and go-to-market motions. Engineers who need a frozen spec before writing code stall; the ones who last prototype against live conversation data, measure AHT reduction or QM automation gains, and iterate. The research team's background (Sebastian Thrun, Tim Shi from OpenAI, Zayd Enam, Ping Wu from Google CCAI) sets a baseline: publishable rigor applied to production deadlines.
Second, they operate with high agency in a remote-first, multi-hub organization. With more than ten hubs across those regions and six hundred employees, there is no central office to absorb coordination overhead. The board's salary data (Staff ML Engineer roles at $230k–$300k, Strategic Sales Directors at $320k–$360k) reflects an expectation that senior contributors drive outcomes without daily stand-ups. Candidates who wait for direction, or who equate "remote" with "async-only," struggle. The culture rewards people who over-communicate context, schedule the sync when threads exceed three replies, and document decisions so the next hub picking up the ticket inherits intent, not just code.
Third, they respect the "human-centric" constraint as a design principle, not marketing copy. Cresta's platform sits inside contact centers where agents handle angry travelers, confused policyholders, and high-stakes financial disputes. The AI Agent and Training Simulator are built to augment — not replace — those workers. Engineers and product managers who optimize for model metrics at the expense of agent trust see their features rejected by customers. The case studies (Oportun cutting QM workload in half, Snap Finance reducing AHT, Brinks Home saving time and cost) all trace back to changes agents actually adopted. People who have never sat beside a contact-center rep, listened to a call, or felt the pressure of a queue tend to build the wrong thing.
Who mismatches? Three patterns appear repeatedly. Specialists who define their role narrowly ("I only do model training," "I only write backend services") collide with a product surface that spans real-time inference, telephony integration, no-code tooling, analytics dashboards, and simulation environments. The interview loop tests for this explicitly: cross-functional exercises, live debugging with product managers, and a values screen that probes whether candidates have shipped something end-to-end without a platform team handing them infrastructure.
Process-oriented operators who equate maturity with bureaucracy also struggle. Cresta's board includes Doug Leone (Sequoia) and Carl Eschenbach (Workday, Sequoia) — investors who back speed over governance theater. The company hit $100M ARR with roughly 775 people; that revenue-per-employee ratio leaves no room for committees that approve committees. Candidates who ask "what's the RACI?" before prototyping a feature tend to self-select out before the offer stage.
Finally, people motivated primarily by research prestige over product impact rarely last. Cresta publishes (patents on conversational AI, Real-Time Trends blog posts analyzing tens of millions of anonymized conversations), but the publication serves the product roadmap, not the reverse. The AI Analyst and Topic Discovery features launched because customers needed earlier warning on emerging issues and a taxonomy they could trust; the research followed the pain. Engineers who want to chase SOTA on public benchmarks without a customer pulling the other end find the feedback loop too tight and the success criteria too commercial.
The filter is functional: Cresta sells to Fortune 500 contact centers that measure success in seconds saved, revenue retained, and agents retained. Every role — ML, backend, sales, marketing, talent — ultimately answers to that scoreboard. If your satisfaction comes from moving that needle, the pace is energizing. If you need insulation from it, the culture will feel abrasive.
Working in frontier tech? Zero G Talent tracks the openings: see every open Cresta role, browse frontier tech jobs, the companies hiring, and the people building the field.