Infrastructure & Security Engineer

San Francisco, CA

About the role

About Engram

Today’s AI is a brilliant stranger: it can solve the world’s hardest math problems, but it knows next to nothing about you and your work. It rereads your files to answer even basic questions, burns an enormous amount of tokens when sifting through large corpuses, and between sessions, it retains scraps at best.

We train models to study your world and anticipate your questions in advance, forming engrams: compact memories that capture your knowledge and history. Our approach opens a new axis of scaling. The more we study your context at training time, the better we become at inference time.

We're already working with leaders in AI like Microsoft, Notion, and Harvey, and just raised $98M from General Catalyst, Kleiner Perkins, Sequoia, Factory, Modern, Amplify, Neo and others. Our investors and advisors include Assaf Rappaport, Andrej Karpathy, and Pieter Abbeel.

AI has spent years learning everything about the world. Now it should learn something about yours.

About the role

Every Engram customer gets models trained on their own context, such as their documents, messages, code, and knowledge bases, and retrained continuously as that context changes. That means isolation can't stop at where data is stored: it has to follow the data into training runs and into the models themselves.

We're looking for an infrastructure engineer to own and evolve the systems that make that true, from ingestion through training to inference, as we add customers and providers. You'll work closely with our researchers so the secure path is also the fast one.

What you'll do

Build and operate the cloud infrastructure responsible for training and serving Engram customer models.
Design and enable the boundaries that keep each customer's data, training runs, and models separate; both on the cloud and on Engram-managed baremetal compute.
Partner with researchers and engineers on paved roads that improve our isolation and security posture.
Evaluate and secure the providers and tools we build on, including Modal, Baseten, and Fireworks as well as compute on baremetal or Kubernetes.
Take the lead on security operations and customer trust: maintain the logging and monitoring we rely on, respond to incidents, and bring clarity to security reviews and compliance work.

Your background looks like

A bachelor’s degree, or equivalent practical experience, in computer science, engineering, machine learning, or a related field.
Four years of work experience as a software engineer or equivalent, exclusive of internships.
Experience building and operating production infrastructure on AWS, including IAM expertise: cross-account roles, STS, resource and key policies, KMS, and multi-account organizations.
Infrastructure-as-Code (Terraform or similar) at multi-account scale, with secure patterns for networks, IAM, and shared services.
A rigorous approach to multi-tenant isolation.
A track record of owning complex, production-critical systems end to end.
Comfort with ambiguity and high ownership at an early-stage company.
Strong production-grade coding skills (Python, Rust, or similar). This role involves shipping code as well as configuration.

No prior ML experience is required.

Bonus points if you have

Experience as an early infrastructure or security hire at a startup.
Experience with ML infrastructure, such as GPU training and model serving.
Experience supporting enterprise security reviews and compliance programs such as SOC 2.
Open-source contributions, talks, or writing on cloud or infrastructure security.
Experience on Azure or GCP.

Engram is based in San Francisco. This role is in-person in our SF office. We offer competitive cash compensation and startup equity.

Engram is an equal opportunity employer. We’re building a team that reflects a range of backgrounds and perspectives, and we welcome applicants regardless of race, color, religion, national origin, gender, gender identity, sexual orientation, age, disability, or veteran status.

About Engram

Today’s AI models don’t understand what you do. Not really. Everything models know comes from their training – and they’re trained mostly on the public internet. They’re knowledgeable about popular Github repos and things people write in articles online. But what you spend your time thinking about every day is so much more than that. You know what good work looks like. You know where a specific project is going, and where you want it to be in a year. You know everything – from the small details, to the big picture of how life and work fit together – knowledge that goes far beyond any chat window. And when you do write things down, important ideas scatter across a sprawl of documents and files. When you use a model, it reads and re-reads many documents from your company before it can even get started. But as we all know by now, this paradigm isn’t perfect: as our contexts grow, models become more expensive and more confused. Individual users already generate far too much data for a model to process. And reading is shallow and temporary; even when the model does see your context, it forgets everything the moment you close the chat. Right now, models do not learn from this data. This means they can’t automatically get better at the things you use them for. We want to change this by building models that learn from your context. Ours is a fundamentally different bet from other labs. Instead of spending massive amounts of training compute on public data, we start from strong pre-trained models and spend training compute on the context you care about. Each model spends the equivalent of hundreds of years studying your context: piecing things together, drawing connections that have never been drawn, finding errors that went unnoticed.