Founding Engineer: ML/ Backend

Before You Apply
You must be able to start now, or within days. We cannot wait out a 60 or 90 day notice period. If you are between roles, on a career break, or currently contracting, we would like to hear from you.

About Memoria
Memoria builds passive in-home sensing for dementia care. Cameras placed in the home run vision models continuously and surface what a caregiver needs to know: whether a parent ate lunch, took their medication, slept through the night, fell, wandered at 2am, or spent the afternoon agitated. No wearables, no panic buttons, and nobody watching a live feed.

We sell to families and to care facilities, and we have letters of intent from home care agencies. The longer-term goal is a health record for the home: a longitudinal model of a person’s routines, preferences, and decline over months.

Our first homes are in Mumbai. India has more than 8 million people living with dementia and almost no formal care infrastructure, which means the burden falls almost entirely on family.

About the Role
You would be our first engineering hire. The centre of gravity is computer vision, backend and systems, but the boundaries are soft. If something needs building and you are the person to build it, it is yours.

The system already runs end to end: edge gateway, cameras, perception pipeline, detectors, and memory layer. You would inherit it and make it hold up across a hundred homes rather than one.

What You Would Own
The system, end to end. In practice:

• Robustness and code quality across a codebase that runs unattended in homes we cannot physically reach.
• Opinions on modelling for our visual inference system. Picking models, picking architectures, offline eval, creating data processing pipelines, opinions on how to make our AI system better
• Telemetry, reporting, and observability. We need to know what every home is doing without asking a customer to check anything.
• Streaming architecture: video from consumer internet connections into inference, reliably and at low cost.
• Pipeline and GPU efficiency. Inference is our dominant variable cost.
• Security, and privacy enforced in the system rather than in a policy document: per-room rules, redaction, consent capture, retention, access control, and our DPDP Act posture alongside counsel.
• Shipping product, not only infrastructure.

What We Are Looking For

• 4+ years shipping backend systems in production. Deep Python and FastAPI.
• Strong experience working with VLMs and traditional computer vision systems. Someone who understands the strengths and limitations of the technology and how to design around it for maximum efficiency.
• Strong PostgreSQL: schema design, migrations, indexing and query performance, row-level security. We run Postgres on Supabase, with pgvector for the memory layer.
• Distributed systems fundamentals: queues, retries, idempotency, backpressure, and partial failure.
• Data pipelines, batch and streaming: schema evolution, backfills, reproducible derived data.
• At least one system you built that ran unattended around the clock with real users on it, and that you were responsible for when it broke. You have done the deployments, the alerting, and the incident response.
• You get real leverage from AI coding tools and review their output carefully.
• Strong written communication. Most decisions here happen in writing.
• Based in India, and able to start immediately.

Helpful, not required: real-time or streaming media (WebRTC, LiveKit, RTSP, ffmpeg), GPU inference and model serving, employee 1 through 5 somewhere before, healthcare or any regulated-data domain.

Rest of the stack: Docker, Railway, per-home gateway appliances, RTSP cameras, LiveKit, APNs, a Next.js web app and a SwiftUI iOS app.

One Thing To Be Clear About
Our product involves video in people’s homes, including bedrooms. We make difficult judgment calls about privacy and dignity regularly. If that sits badly with you, this is not the right role.

How We Work
I am based in California and in India roughly every twelve weeks, more often when the work demands it, including for customer installations. Our fixed overlap window is 8pm to 10pm IST. Everything outside that happens asynchronously in writing.

Compensation
Contract: Up to INR 2,000 per hour for the first three months, invoiced monthly. You would handle your own tax filings and GST where applicable.

On conversion: We can formally negotiate salary and stock options upon the conclusion of the 3 month trial.

We intend to convert this role to full-time, and that decision is tied to closing our current round. We will tell you where the raise stands on the first call and again at the end of the contract, either way.
Equity is granted at conversion. We are a Delaware C-corp, so you would receive stock options under our US plan. Indian residents are taxed at exercise, on the difference between strike price and fair market value, even without a sale. We will explain your exposure in numbers before you sign. There are no buyback restrictions or forfeiture beyond standard vesting.

We are pre-seed and actively raising, and we will give you our runway number on the first call. Hardware, tools, and travel are covered.

Hiring Process
Under two weeks.
90 minutes with the founder: the company, the role, a system you have shipped, and an architecture discussion about our infrastructure.
A paid two week trial project on a real problem from our backlog, at your day rate.
References, then an offer.

How to Apply
Send us:

1. Your earliest possible start date, in the first line.
2. A GitHub profile, or code you can share.
3. One system you built that ran in production, what broke, and how you fixed it.
4. Whether you are currently under any employment or contracting obligation.