Member of Technical Staff [AI/ML Engineer]

theburntapp.com•San Francisco, CA

1d•Onsite

About The Position

Burnt is building the Operating Brain for the global supply chain, a living, evolving brain that will run businesses on autopilot. The brain runs on models, transforming historical data from ERPs into forecasts, an ontology for agents to reason over, and models tuned on proprietary data. The company is starting in the food industry, a complex and unforgiving sector. Burnt is culturally competitive, aiming to build the future of supply chain operations. This role is for the first dedicated ML hire, responsible for building the model layer. The focus is on improving system performance, enabling continuous learning, and achieving scalable margins. Key challenges include demand forecasting, developing an ontology and knowledge graph for agent reasoning, and bringing LLM fine-tuning in-house. This is a production-focused role, not a research position, requiring the ability to architect systems, defend design decisions, and contribute to full-stack development when necessary. The team is small and ships fast, valuing individuals who proactively take on tasks outside their immediate scope.

Requirements

Python at expert level.
Apache Spark and the surrounding data engineering ecosystem.
AWS SageMaker as the core platform, plus S3, Glue, Step Functions and the rest.
MLflow, Kubeflow, or equivalent for MLOps.
LoRA and PEFT libraries such as HuggingFace PEFT or TRL for fine-tuning.
Prophet, NeuralForecast, statsmodels, or similar for forecasting.
Neo4j, RDF, OWL, SPARQL, or similar for knowledge graph.
TypeScript and React for the application layer.
Demonstrable expert-level Python skills.
Experience deploying and maintaining ML models (not agents) in production.
Hands-on experience with AWS SageMaker.
Experience with a time series forecasting model in production.
Experience with LLM fine-tuning using LoRA or PEFT.
Experience with ontology or knowledge-graph-backed systems in a real product context.
Willingness and ability to pick up full stack work when the team needs it.

Nice To Haves

Ability to ramp fast on technologies not explicitly listed.
Ability to ship a frontend.

Responsibilities

Own MLOps end to end: model creation, deployment, iteration, monitoring, and support.
Build and maintain time series forecasting models that serve production traffic and hold up under backtest against real order books.
Fine-tune LLMs with LoRA and PEFT on our own data, and build the evals that decide what ships.
Design and maintain the ontology and knowledge graph our agents reason over.
Engineer data pipelines at scale with Spark, over messy multi-tenant supply chain data.
Build the versioning, drift detection, and retraining pipelines that keep models honest after launch.
Run the AWS ML stack: SageMaker at the core, with S3, Glue, and Step Functions around it.
Architect the systems your models live inside, not just the models, and own those design decisions.
Step into full stack work when the team needs it, from the API that serves a prediction to the interface a buyer actually uses.