Applied AI Engineer, Medal.tv

• We're looking for a technical polyglot
• You might have started in systems engineering, physics, or neuroscience and moved to ML, or the other way around
• You know Python and PyTorch, but you aren't afraid of C++ or low-level hardware constraints
• 5+ years experience taking complex systems from prototype to production, within software engineering or applied AI/ML
• Strong experience in the ML stack (Python, Docker, Kubernetes, infrastructure-as-code, and CI/CD for ML pipelines) with competent systems programming skills (C++, Go, Rust, or Java), and ability to use modern AI coding tools
• Strong applied machine learning experience, specifically in the lifecycle of deploying, evaluating, and debugging models
• Experience in at least one of the following, with working knowledge of the others:
• Agents or policy learning (e.g., RL, planning, control theory, spatial reasoning)
• World models, simulation environments (Unity/Unreal, Omniverse, Isaac Sim), or model-based learning
• Perception, sensor fusion, or inverse dynamics models (IDMs)
• Exposure to bridging the "hardware-software" gap: integrating AI inference with sensors, edge devices, RTOS, or legacy industrial networks
• Full-stack systems mindset: understanding of memory management, concurrency, networking, and APIs
• U.S. citizenship and ability to obtain and maintain a national security clearance (TS/SCI preferred)
• Ability to comply with export control requirements (ITAR/EAR)

Desirable

• We are looking for an Applied AI Engineer to connect our research with the reality of our partners’ environments that are constrained by hardware, power, and real world interference
• We work with customers operating in complex areas across robotics, simulation, aerospace and defense, manufacturing, logistics, industrial automation, and more
• You will be responsible for taking our model, focusing on post-training, evaluation data, and integrations to ensure our customer's platforms work regardless of a messy, constrained tech stack and hardware
• You will embed with our partners to understand their actual problems, not just what they put in an RFP
• You’ll look at their legacy control systems, latency challenges, and power needs and figure out how our AI helps them achieve something that was never possible before
• You'll be a key part of the feedback loop
• When a model fails because of sensor noise or unexpected physics, you don't just log a bug
• You figure out why, and you work across our team to fix the underlying architecture
• You ensure we are building technology that survives contact with the real world, for years to come
• Most importantly, you have high agency and want to be a part of an amazing team
• Embed with partners to solve the problems with our frontier AI/ML tools, informing our research and product development plan along the way…not just deploy software
• Be the primary filter between the messy reality of the physical world and our research and technical staff, surfacing real commercial challenges and pain-points
• Build and tune models, prototype, script, and patch (often in the field), turning ambiguous requirements into executable code
• Build the connective tissue between our AI and the customer's reality, and then help them rethink the art of the possible. This means writing high-performance code (C++/Go/Python) that integrates our inference engine with legacy sensors, RTOS, and diverse hardware peripherals
• Optimize complex ML models for survival in harsh computing environments
• Leverage our simulation and world-model capabilities to validate operational plans before they touch physical hardware
• Translate the probabilistic nature of AI into the deterministic language of industrial control systems and mission operators
• Explain trade-offs to non-technical individuals and deep technical details to systems engineers, building the trust required to deploy autonomous systems in critical paths
• The Stack:
• ML & Research: Python, PyTorch, NumPy, OpenCV, Triton, CUDA for large-scale training, real-time inference, and applied CV/ML
• Pipelines & Experimentation: Kubeflow Pipelines and Airflow with continuous evaluation, A/B testing, and performance monitoring across training and production
• Backend & Systems: Java services with Redis and RabbitMQ, plus performance-critical C++ components; containerized with Docker and Kubernetes on GCP or on-prem
• Clients & Edge Software: Electron/React desktop apps, C# and C++ high-performance recorders, and mobile clients in Swift (iOS) and Kotlin (Android)
• Infra, Hardware & Other Deployments: Terraform-managed infrastructure, CI/CD via GitHub Actions and CircleCI; deployment to NVIDIA GPU clusters, air-gapped or on-prem environments, hardened Linux systems (FIPS/STIG), and constrained real-world hardware requiring model optimization, hardware-specific acceleration, and secure supply-chain practices