Senior Data & ML Infrastructure Engineer (Xora Portfolio Company)

Xora Innovation; San Diego (CA); Hybride; USD 130 000 - 180 000; Plein temps

Résumé du poste

ELEMYNT, built by Xora Innovation, seeks an experienced systems engineer to own data and ML infrastructure. You will architect data pipelines, curate training data, and manage model packaging, serving, and CI/CD across on-site or hybrid environments.

You will ensure observability, reliability, and performance of deployments in customer environments, including cloud and HPC setups, with multi-GPU support and scalable workflows.

Qualifications

• 6+ years shipping production software with data systems and ML infra.
• Strong Python and end-to-end system delivery experience.
• Experience with large-scale data systems and distributed compute.
• Production MLOps: packaging, versioning, monitoring, CI/CD.

Responsabilités

• Build and operate data pipelines for large-scale scientific output.
• Develop ML data pipelines for training, fine-tuning, RL.
• Package, version, deploy models and maintain CI/CD.
• Run CI/CD and serving for batch, online, and async inference.
• Monitor drift, latency, and anomalies; set up dashboards and alerts.
• Design APIs and internal tools to improve reliability for engineers.

Connaissances

Python

Data pipelines

MLOps

Containers & orchestration

Cloud & HPC

Formation

BSc or MSc in CS or related

Outils

Airflow

Dagster

Flyte

Kubernetes

OpenTelemetry

Prometheus

Description du poste

About Elemynt

ELEMYNT is an early-stage startup built by Xora Innovation. We develop applied intelligence that brings AI into the real world. Our platform combines advanced machine learning, high-performance simulation, and modern software engineering to accelerate the design, validation, and deployment of new materials. Our work sits at the intersection of AI, physics, and large-scale computation. The problems are hard, the stakes are high, and the impact is tangible.

ELEMYNT is an early-stage startup built by Xora Innovation. We develop applied intelligence that brings AI into the real world. Our platform combines advanced machine learning, high-performance simulation, and modern software engineering to accelerate the design, validation, and deployment of new materials. Our work sits at the intersection of AI, physics, and large-scale computation. The problems are hard, the stakes are high, and the impact is tangible.

About The Role

This role builds and operates the data and machine-learning infrastructure the platform runs on: the pipelines that turn large-scale scientific output into data models can train on, and the systems that move those models from research into production and keep them running there. Hands-on work, close to both the data and the models. You'll own both sides. On the data side, that's pipelines and formats that keep large-scale output fast to query and ready for training. On the model side, it's the packaging, serving, monitoring, and CI/CD that let models ship safely and stay healthy once they're live. And because the platform runs inside customers' own secure environments, on their clusters, in their cloud, or a mix of the two, whatever you build has to stay observable and reliable in places you don't operate. Everything downstream depends on this layer. When it's slow or unreliable, so is everything built on top of it.

What You Will Do

• Build the data pipelines that ingest, transform, and curate large-scale scientific output into efficient, training-ready formats on object storage.
• Make that data fast to query and cheap to reuse, so analysis and downstream jobs aren't left waiting on it.
• Build the ML data pipelines for training, fine-tuning, and reinforcement learning: curation, deduplication, formatting, and the evaluation sets that keep training honest.
• Catch bad data early, with validation and quality gates that check schema, distribution, and completeness before it reaches a model.
• Package, version, and deploy models across development, staging, and production, with registries and reproducible builds that keep every deployment traceable.
• Run models through CI/CD and serving workflows for batch, online, and asynchronous inference, with safe rollout, rollback, and quick diagnosis when something breaks.
• Monitor deployed models for drift, degradation, latency, and anomalies, with automated regression checks that flag trouble before users do.
• Stand up dashboards, metrics, logs, and alerts that surface data and model problems while they're still small.
• Design the APIs, services, and internal tools that make these workflows reliable and easy for engineers and scientists to use.

What We Are Looking For

• Bachelor's or Master's degree in Computer Science or a related engineering field, and 6+ years building and shipping production software, with real depth across data systems and ML infrastructure.
• Strong Python, and a track record of shipping reliable systems end to end that other people end up depending on.
• Hands-on experience with large-scale data systems: object storage, efficient columnar and array data formats, and distributed query and compute engines.
• Experience building data and ML data pipelines: ingestion, transformation, curation, and the validation and quality gates that catch problems before they reach training or inference.
• Production MLOps experience: packaging, versioning, serving, and monitoring models for drift, latency, and anomalies, backed by model registries and CI/CD for ML.
• Deep hands-on experience with containers and orchestration (Docker, Kubernetes) and workflow orchestrators such as Airflow, Dagster, Flyte, or Temporal.
• Experience instrumenting production systems and using their telemetry, logs, and metrics (Prometheus, Grafana, OpenTelemetry, or similar) to debug real incidents.
• Comfort working across cloud and HPC, including distributed multi-GPU, and owning ambiguous systems end to end in an early-stage setting with little scaffolding.

NICE TO HAVE

• Data-quality and governance tooling such as Great Expectations or Evidently, plus data contracts, lineage, metadata catalogs, or reproducibility tooling.
• Model-serving patterns for high-throughput or asynchronous inference, and runtime uncertainty or out-of-distribution monitoring.
• Experiment-tracking and model-lifecycle tooling such as MLflow or Weights & Biases, or serving stacks such as Ray Serve, KServe, or Kubeflow.
• Experience applying ML to scientific data, such as property prediction, generative models, or graph-based approaches.
• It'd be a plus if you've worked with atomistic-ML data tooling: Atompack, ASE-style structure databases, extended-XYZ datasets, or the large public corpora built on them.
• Contributions to open-source ML or data infrastructure.

LOCATION

Singapore or United States. We're hiring in both to reach the right person. Work model is on-site or hybrid, set per location.

CLOSING NOTE

If you don't tick every box but this is clearly your kind of work, get in touch.