Machine Learning Engineer

Secludy, Inc.; San Francisco (CA); Hybrid; USD 140,000 - 210,000; Full time

Job summary

Secludy, Inc. in San Francisco is building a privacy-preserving AI platform and seeks an ML engineer to join as the first engineering hire at a seed-stage startup.

You will turn research into production-ready infrastructure for synthetic data with provable privacy guarantees, partnering with founders to ship high-impact solutions. You will optimize distributed training on Kubernetes, own the deployment pipeline, and drive performance while debugging complex failures.

Qualifications

• 3 to 6 years in ML engineering or applied research.
• Strong PyTorch experience and ability to write a custom training loop.
• Experience with generative models (diffusion, VAEs, transformers) beyond inference.
• Track record of taking experimental code to production.
• Comfort with tabular, text, or time-series data.

Responsibilities

• Accelerate research loop by turning mathematical hypotheses into testable code.
• Productionize the engine and own the deployment pipeline.
• Orchestrate distributed training and generation across GPU clusters with Kubernetes.
• Build logging that explains why a generation job failed.
• Diagnose failures spanning infrastructure and model training.

Skills

PyTorch

Custom training loop

Gen models (diffusion/VAEs/transformer

Production deployment

Diagnostics & debugging

Tools

Kubernetes

Docker

Job description

Build the infrastructure for safe enterprise AI

Join a seed-stage team making privacy-preserving AI the default. We help enterprises ship AI when sensitive data is the blocker. Your work turns restricted datasets into synthetic data with provable privacy guarantees that AI, legal, security, and compliance teams all approve.

We are building a proprietary generative engine that creates privacy-protected synthetic data for regulated industries like fintech, healthcare, and life sciences. Small team, high ownership, direct work with the founders.

Founded in San Francisco

Hybrid in the SF Bay Area

Venture-backed and seed-stage

High ownership

Small team, real scope. You own problems end to end and ship work the founders and customers feel right away.

Pragmatism

Write clean abstractions when they pay off. Ship a small script when the team needs an answer today.

Privacy-first

We make individual privacy hold as AI scales. Provable privacy guarantees are the product, not an afterthought.

Speed

We turn math hypotheses into testable code fast and keep the loop between research and production tight.

Secludy is a San Francisco-based, venture-backed startup on a mission to make privacy-preserving AI the industry default and keep individual privacy intact as AI scales. The platform lets organizations use sensitive data through synthetic data generation with provable privacy guarantees.

You will be the first engineering hire at a seed-stage startup solving the data bottleneck for regulated industries. You will build the infrastructure that lets the team iterate much faster and deliver to enterprise customers at a high standard.

What you will do

Accelerate the research loop

• Partner with the founders to turn mathematical hypotheses into testable code.
• Build reusable modules and keep the codebase simple by questioning every new file and library.
• Push performance through simpler architecture, not patches.
• Investigate failures that span infrastructure and math, like out-of-memory errors, race conditions, loss divergence, and gradient explosion.
• Build the harness that shows exactly how metrics change when the model changes.

Productionize the engine

• Orchestrate distributed training and generation across GPU clusters with Kubernetes using minimal, self-healing deployments.
• Own the deployment pipeline.
• Build logging that explains why a generation job failed, not just that it failed.

Who you are

• You work across infra and modeling. You can debug latency and Docker issues, read loss curves, and diagnose overfitting.
• You take ownership of broken training runs. You profile GPUs, inspect dataloaders, and find the bottleneck fast.
• You stay pragmatic. You write clean abstractions when they pay off and ship a small script when the team needs an answer today.

Requirements

• 3 to 6 years in ML engineering or applied research.
• Strong PyTorch. You can write a custom training loop from scratch.
• Experience with generative models like diffusion, VAEs, and transformers beyond inference.
• A track record of taking experimental code to production.
• Comfortable with tabular, free-text, or time-series data.

1. Screen

30 minutes

Align on mission and background.

60 minutes

ML system design, debugging war stories, and generative concepts.

3. Working session

90 minutes

Diagnose and fix a real broken piece of research code in your own environment while we collaborate.

4. Founder chat

30 minutes

Final vision alignment.