Machine Learning Engineer

LiquidXR is seeking a Machine Learning Engineer to help build advanced models that turn multimodal time-series sensor data into meaningful signals. In this hybrid role based in Los Angeles, you will focus on robust real-time algorithms that perform reliably with noisy, high-frequency inputs.

The work spans modeling, training strategy, and practical deployment, from research-style experimentation through to production-grade inference under latency and compute constraints.

What you’ll do

• Design and implement machine learning models for time-series and sequential data
• Develop algorithms to extract structured signals and latent variables from noisy sensor inputs
• Build and optimize real-time inference pipelines with latency and compute constraints
• Apply multi-modal learning and sensor fusion techniques
• Replace or augment classical signal processing pipelines with learned models
• Design training strategies for windowed and streaming data
• Design training strategies for weakly labeled or partially observed datasets
• Design training strategies for multi-task learning setups
• Evaluate models using both statistical metrics and application-driven performance criteria
• Collaborate with cross-functional teams to take models from research to production
• Explore architectures including Temporal Convolutional Networks (TCNs)
• Explore architectures including RNNs, LSTMs, and GRUs
• Explore architectures including Transformer-based sequence models

What you bring

• Strong experience with machine learning for time-series data
• Experience with transfer learning and knowledge distillation techniques
• Proficiency in Python and PyTorch (or similar frameworks)
• Solid understanding of signal processing fundamentals including filtering, noise, and the frequency domain
• Experience working with real-world, noisy datasets
• Experience building or deploying low-latency or real-time systems
• Experience with sensor data such as IMUs
• Familiarity with sensor fusion methods such as Kalman filters and probabilistic models
• Experience with multi-modal or multi-task learning
• Exposure to embedded or edge deployment constraints
• Background in applied domains involving physical systems or human data
• BSc or MSc in quantitative fields such as computer science, engineering, physics, or applied math
• Ability to reason about temporal structure, causality, and latency
• Strong intuition for modeling tradeoffs versus deployment constraints
• Comfort working with imperfect, real-world data
• End-to-end ownership from modeling through validation to deployment
• Self-starter mindset with the ability to catch the vision and run with it
• Clear communicator and effective collaborator with cross-functional teams
• Ability to manage multiple priorities without sacrificing quality
• Detail-oriented approach focused on high-quality and well-documented work
• Ownership mindset with full accountability from concept to completion

Tools and technologies

• Python
• PyTorch
• RNNs
• LSTMs
• GRUs
• Transformer-based sequence models
• Temporal convolutional networks (TCNs)
• Kalman filters

Benefits

• Employee stock option program
• Health care benefits: currently, gold PPO coverage with Blue Shield, plus dental and vision, starting within 30 days of employment
• Open PTO company

Location: Los Angeles, CA (hybrid). Employment: Full-time employee position, working remotely or in the Los Angeles office. Compensation: Commensurate with experience and competitive with the market. Travel: Occasional domestic and international travel may be required.