Job description

About Mappa

Mappa builds real-time voice-AI infrastructure that decodes over 100,000 behavior patterns per person—based on prosodic and linguistic signals—and turns them into actionable insights for hiring and beyond. We’re growing fast and building novel LLM-based systems that power behavior interpretation at scale.

Role Purpose

Design, implement, and maintain agentic systems that turn raw voice data into intelligent decisions. You'll own the architecture and prompt logic behind Mappa’s LLM pipelines—ensuring safe, scalable, and high-precision outcomes.

Key Outcomes (First 6 Months)

• Build and deploy core LLM agents capable of executing multi-step reasoning tasks with memory, safety checks, and structured handoffs.
• Deliver prompt architectures that support real-time interaction with prosodic and linguistic features.
• Implement testing and evaluation workflows to monitor hallucinations, latency, and performance of agentic flows.
• Collaborate with behavioral scientists and product teams to continuously refine the signal-to-insight loop.

Core Responsibilities

• Architect and maintain prompt frameworks, chains, and guardrails for LLM-based behavior inference.
• Build custom agents using tools like OpenAI’s Agents Framework, integrating with internal APIs and voice-processing layers.
• Implement memory systems, feedback loops, and handoff logic to other systems (human or machine).
• Ensure safety, reliability, and transparency across LLM interactions.
• Partner with Engineering and Data Science to continuously test, fine-tune, and improve LLM performance.
• Document systems, prompts, and experiments for internal reproducibility and iteration.

Must-Have Skills

• 3–5 years of experience in backend or ML engineering with exposure to LLMs and agentic architectures.
• Strong knowledge of Python and/or TypeScript.
• Hands-on experience with prompt engineering and LLM APIs (OpenAI, Anthropic, etc.).
• Familiarity with agent frameworks (e.g., OpenAI Agents, LangChain, or similar).
• Understanding of memory management, handoff logic, and prompt evaluation techniques.
• Ability to collaborate closely with product, behavior science, and data teams.
• Fluent English.

Nice-to-Have

• Experience working with safety tooling for LLMs (e.g., moderation APIs, rule-based filters, evaluation tooling).
• Exposure to voice, audio, or NLP pipelines.
• Familiarity with real-time or latency-sensitive AI systems.

Who You Are

Builder of smart systems. You move fast, test relentlessly, and are passionate about using LLMs not just to chat—but to reason, decide, and act. You care about edge cases, safety, and crafting behavior that feels intentional and reliable.

Why Join Mappa

• Deep tech: Work at the intersection of LLMs, voice, and behavioral science.
• High ownership: Build the agentic brain behind our core products.
• Remote-first: Join a distributed team across LatAm and the U.S.