Founding Software Engineer, Platform & Infrastructure - DeepFrame - at Infinimind

Responsibilities

You will build and scale GPU-based infrastructure for video understanding, including ingestion, indexing, and search pipelines. Additionally, you will collaborate with the GTM lead to support US enterprise customers and integrate their requirements into the core product.

Requirements

Candidates must have 5+ years of experience in production infrastructure or backend platforms with deep knowledge of distributed systems. Proficiency in Python and a systems language like Go, Rust, or C++ is required, along with experience managing GPU workloads and large-scale data pipelines.

Skills

• Python
• Go
• Rust
• C++
• AWS
• Kubernetes
• GPU Infrastructure
• Distributed Systems
• Model Inference
• ML Pipelines
• Data Pipelines
• Video Processing
• Search Infrastructure
• API Development
• System Architecture
• Enterprise Deployment

About this role

WHAT DEEPFRAME IS: 

DeepFrame is InfiniMind's video intelligence platform. Enterprises point it at
video they already have - broadcast programming, in-store footage, operations
video - and get evidence-backed answers: search across thousands of hours,
questions answered with timestamped clips, and rulebook reviews that turn a
company's own standards into findings with severity and proof.
 
We are not an AI wrapper. We build the full stack ourselves: our own
video-understanding foundation models, ingestion and indexing across visual,
speech, on-screen text, and time, a search engine, and agents on top. It runs as
SaaS or single-tenant in the customer's own AWS account, with on-prem options
for footage that cannot leave the building. 

DeepFrame has paying enterprise customers in Japan in broadcasting and retail.
You will be the first engineer on the ground in the US, working with a US GTM
lead hired in parallel and with our core engineering and research team in
Tokyo. 

WHAT THIS ROLE REALLY IS: 

An infrastructure and platform engineer with a feel for AI - not an "AI
engineer" who wraps someone else's API. Roughly: 
* ~75% building the platform - GPU inference and video-processing
infrastructure, ingestion and indexing pipelines, search and serving,
performance and cost at scale, shipped into the shared DeepFrame codebase
with the Tokyo team 
* ~25% with customers - joining technical meetings with the GTM lead, answering
integration and security questions, and getting pilots running on the
customer's own video 

When a US customer needs something, you build it properly in the core product,
not as a one-off fork.

HOW THE US TEAM WORKS:

You and the US founding GTM lead are DeepFrame's US team. They lead the customer
conversation; you join technical discovery and make sure what we commit to can
be built and run. 
* Daily with the CTO on the technical side - architecture, priorities, code
review, and how US needs land in the core product with the Tokyo engineering
team 
* Daily with the CEO on the business side - which customers and pilots matter
most, and what we commit to 
* A short daily sync across Palo Alto and Tokyo (late afternoon Pacific /
morning Japan), plus a weekly US pipeline and pilot review 
* You report to the CEO. The CTO owns architecture, technical direction, and
code review; when commercial and technical priorities conflict, the CEO
decides what we commit to a customer, with the CTO owning technical
feasibility 

WHAT YOU WILL DO:

* Build and run GPU infrastructure for video understanding on AWS: model
serving, batching, scheduling, and cost per video-hour 
* Scale ingestion and indexing pipelines for large video libraries: decoding,
transcoding, frame, audio, and on-screen text processing, embeddings 
* Work on the search and retrieval layer and the APIs customers integrate with 
* Build connectors to where US customers keep video (S3, media asset management
systems, camera exports) and to where they want results (APIs, webhooks) 
* Deploy for enterprise customers - SaaS, single-tenant in their AWS account,
or on-prem - and own reliability and observability 
* Join customer technical meetings with the GTM lead and turn what you hear
into product requirements 

YOUR FIRST 90 DAYS:

* Day 30 - DeepFrame running in a test AWS environment. First production change
merged. A measured view of where our GPU and pipeline costs go. 
* Day 60 - One infrastructure improvement owned end to end (throughput, cost,
or reliability). Pilot environment ready for the first US customer. 
* Day 90 - Technical owner of the first US pilot, from data access through
first results review. At least one US-driven feature or connector shipped to
the core product. 

MUST HAVES:

* 5+ years building production infrastructure or backend platforms, with real
distributed-systems depth 
* Strong Python, plus a systems language (Go, Rust, or C++) 
* GPU workloads in production - you have run and optimized model inference or
ML pipelines on cloud GPUs (CUDA, TensorRT, Triton, or similar); AWS
preferred; containers and Kubernetes
* Large data or media pipelines - you have moved and processed a lot of data
reliably and cheaply
* Can explain technical tradeoffs to a customer's engineers in a meeting 
* Authorized to work in the United States by your start date; we do not sponsor
visas for this role
* Based in the Bay Area, in person in Palo Alto at least 3 days a week, and
able to travel to customers occasionally 

STRONG PLUSES: 

* Video and media processing (ffmpeg, codecs, GStreamer, NVIDIA DeepStream /
NVDEC)
* Search infrastructure: vector databases, retrieval, and ranking at scale 
* ML systems: serving foundation models, evaluation pipelines 
* Enterprise deployment: VPC / on-prem installs, SSO, security reviews 
* Worked with a team across time zones, especially Japan; working proficiency
in Japanese is a plus, not required 

About Cox Exponential

Technology, Information and Media · 10,001+ employees · headquartered in Atlanta, Georgia · founded 1898