Build the Decision Plane for Consequential Execution

Memrail is building SOMA AMI - the decision plane for autonomous decisions. Deterministic context assembly, policy gates, and auditable traces. We're hiring engineers who want to make autonomous behavior reproducible in production.

4
Open roles
Miami
Florida

What We're Building

Decision Plane Runtime

EMU runtime, trigger reachability, policy compilation, deterministic activation, idempotency.

SDK & Integrations

Python-first SDK plus integrations for LangGraph, n8n, LlamaIndex, and custom orchestrators.

Tracing & Replay

Decision traces, diffable replays, and debugging tools for cognitive systems in production.

Open Positions

ML Ops Engineer

LLM Training RLHF/RLAF MLOps Python
Miami, FL · Contract / Full-time +

Own the ML infrastructure and training pipelines that power SOMA AMI - from LLM fine-tuning to reinforcement learning from human and AI feedback. You'll build the systems that make our models learn, adapt, and improve at scale.

What You'll Do

  • — Design and operate LLM training and fine-tuning pipelines
  • — Implement RLHF and RLAF workflows for model alignment
  • — Build RL-based systems for policy optimization and agent learning
  • — Optimize training infrastructure for cost and performance
  • — Design evaluation frameworks and model quality metrics
  • — Maintain model versioning, serving, and A/B testing infrastructure

Requirements

  • — Experience training and fine-tuning LLMs at scale
  • — Hands-on experience with RLHF, RLAF, or similar alignment techniques
  • — Strong reinforcement learning background (PPO, DPO, reward modeling)
  • — Proficiency in Python and ML frameworks (PyTorch, JAX, HuggingFace)
  • — Experience with distributed training and GPU infrastructure
  • — Familiarity with MLOps practices (experiment tracking, model registry)
Apply Now

Cognition Engineer — Agent Architecture

Python LLM Systems State Machines
Miami, FL · Contract / Full-time +

Own agent architecture on top of Memrail's decision plane -- designing how agents perceive, deliberate, and act under deterministic context + policy constraints. You'll sit at the intersection of LLM agents, systems design, and cognitive architectures.

What You'll Do

  • — Design agent control flows (planning, reflection, tool use, recovery)
  • — Define cognition patterns for real use cases
  • — Architect EMU encoding, recall, and activation
  • — Prototype meta-cognitive loops
  • — Translate fuzzy requirements into cognitive contracts
  • — Document reusable cognition templates

Requirements

  • — Experience shipping LLM-powered systems or agents
  • — Solid backend/systems engineering (Python)
  • — Familiarity with RAG, vector DBs, knowledge graphs
  • — Comfortable with state machines, graphs, workflows
  • — Clear reasoning about failure modes and safety
Apply Now

Cognition Engineer — Memory Systems & EMUs

Data Pipelines Knowledge Graphs Symbolic AI
Miami, FL · Contract / Full-time +

Build the substrate layer of artificial cognition: how state/events become executable memory (EMUs), how triggers stay reachable, and how behavior stays auditable over time. You'll work directly with SOMA AMI and Executable Memory Units (EMUs) to design deterministic, auditable memory for AI agents.

What You'll Do

  • — Design memory schemas and EMU taxonomies
  • — Define symbolic triggers for EMU activation
  • — Work on reachability and connectivity
  • — Build pipelines: logs → fragments → EMUs → meta-EMUs
  • — Evaluate memory quality with precision/recall metrics
  • — Document memory patterns and validation procedures

Requirements

  • — Experience with schemas, ontologies, knowledge graphs
  • — Strong Python and pipeline skills (ETL, streaming)
  • — Familiarity with RAG, triple stores, or rule systems
  • — Ability to reason about determinism
  • — Clear technical documentation skills
Apply Now

Cognition Engineer — Cognitive Runtime Operations

SRE Observability Kubernetes
Miami, FL · Contract / Full-time +

Own the runtime behavior of cognition: how EMUs execute, how tools are invoked, how latency/cost are controlled, and how incidents are detected and mitigated. Part SRE for cognition, part runtime architect.

What You'll Do

  • — Operate and tune SOMA/Memrail for production
  • — Define guardrails, timeouts, safety constraints
  • — Build dashboards for EMU activations and decision traces
  • — Fix runaway loops and unexpected EMU chains
  • — Participate in post-mortems (system + cognition)
  • — Define operational playbooks for client deployments

Requirements

  • — Strong production systems experience (SRE/DevOps)
  • — Comfortable with logs, metrics, traces (OTel, Prometheus)
  • — Experience running LLM systems at scale
  • — Ability to reason about behavior over time
  • — Python, Docker, Kubernetes
Apply Now

Memrail is part of the Cadenzai family - research (Cadenzai), infrastructure (Memrail), and applied practice (CognitionEngineers). Roles collaborate across entities.

Cadenzai — Research roles CognitionEngineers — Applied roles

Don't see a role that fits? We're always looking for exceptional people.

careers@memrail.com