Staff Software Engineer - Agent Architecture

PayNearMe is a payments technology company on a mission to simplify how people pay and how businesses get paid. The PayXM platform manages the full payment experience across cards, ACH, digital wallets, and cash at 62,000+ retail locations, serving businesses in consumer lending, iGaming, online sports betting, property management, and tolling. The company raised a $50M Series E in September 2025 and employs 300+ people across 41 states, headquartered in Silicon Valley with offices in Dallas, TX and Holmdel, NJ.

We build agentic AI products that our customers interact with across different modalities. These agents sit on top of the same money-movement platform that handles real funds for businesses in regulated industries, so they must meet safety, compliance, and predictability standards that exceed typical consumer AI products.

Responsibilities:

• Define architectural direction for agentic AI in partnership with engineering leadership. Develop patterns for a multi-tenant agent platform where business customers have different rules, brand voices, allowed actions, knowledge bases, and compliance requirements. Maintain architecture documentation (current state, target state, migration plans) and drive alignment across product, engineering, security, and compliance.
• Design, build, and ship production agents including voice and chat agents for payment-related activities that integrate with the Ruby on Rails/MySQL platform and partner services like ElevenLabs and Twilio. Treat tool design as foundational: tool schemas, descriptions, idempotency, side-effect semantics, and error surfaces directly determine agent quality and accountability for money-moving actions.
• Make and defend decisions about what kind of intelligence goes where: partner stacks vs frontier LLMs, RAG vs tool calls vs fine-tuning, small/fast/cheap models vs frontier models, and where classical ML or deterministic logic outperforms LLMs. Design integration points that allow provider, voice vendor, and model changes without rewriting agents.
• Design and operate the agent lifecycle as a closed loop: testing, offline and online evaluations, observability, scoring, and disciplined feedback paths from production signals to test suites and evaluation sets. Own rollout discipline for non-deterministic systems including prompt and agent versioning, shadow mode, canary-by-tenant deployment, gradual ramps, and rollback playbooks.
• Own the unit economics of agent interactions through token budgets, prompt and semantic caching, model cascades, batch APIs, and latency-versus-cost tradeoffs, with per-tenant cost attribution.
• Build guardrails that make agents safe in payments contexts: scope enforcement, refusal behaviors, deterministic handoffs for money-changing actions, PCI-compliant card data handling, PII protection, and human-in-the-loop fallback paths. Own identity and consent models for agent-initiated actions including step-up authentication. Treat prompt injection and agent social engineering as real attack surfaces through continuous red-teaming.
• Treat voice as a distinct modality, designing for latency budgets, barge-in and turn-taking, STT/TTS error modes, DTMF fallback, recording and consent, and telephony realities.
• Partner with Security, Compliance, and Legal to ensure agent behavior meets PCI-DSS, state-level payments regulations, and customer compliance obligations. Make agent decisions reconstructable for regulators, auditors, and disputing parties.
• Raise the engineering bar for agent work across the organization by defining shared patterns for prompts, tools, evaluations, telemetry, and incident response, and teaching the team how to build, evaluate, and operate agents.

Minimum Requirements:

• 8+ years of software engineering experience with Staff-level scope (cross-team influence, major initiatives, long-term technical direction).
• Demonstrated experience shipping agentic AI systems to production (not prototypes, not internal copilots, but agents that real users have relied on).
• Hands-on experience with at least one modern agent framework like LangGraph or comparable, with understanding of tradeoffs between graph-based orchestration, ReAct-style loops, and deterministic state machines.
• Deep, lived experience with the full agent lifecycle: prompt and tool design, offline and online evaluation, scoring rubrics, observability and tracing, and feeding production signals back into evaluation sets.
• Strong system design fundamentals (reliability, consistency, data modeling, pragmatic API/service boundaries). Ability to integrate agents into existing transactional systems.
• Comfortable in a Ruby on Rails/MySQL environment or able to ramp quickly.
• Clear communication and strong judgment in high-stakes cross-functional environments.

Preferred Qualifications:

• Experience shipping voice agents (latency, turn-taking, STT/TTS failure modes, prosody, barge-in, DTMF, telephony, recording/consent).
• Hands-on experience with voice-enabled applications using platforms like ElevenLabs, Twilio, Vonage, Deepgram, Vapi, or LiveKit.
• Experience designing agents combining multiple intelligence styles (partner-managed conversational stacks, frontier LLMs, smaller models, RAG, classical ML, deterministic logic).
• Payments/fintech experience or other regulated/high-integrity domains (healthcare, lending, insurance). PCI-DSS or TCPA experience strongly preferred.
• Experience building evaluation infrastructure (LLM-as-judge, rubric-based scoring, regression suites, A/B and shadow testing).
• Experience designing multi-tenant agent platforms and adversarially testing LLM systems (prompt injection, jailbreaking, social engineering of tool-using agents).

Compensation: $225,000 - $285,000 USD base, plus stock options (1-year cliff, 4-year vesting), premium medical/vision/dental (100% employee coverage), 401k, flexible time off, and home-office stipends.