Perception Engineer

Brahma Consulting Group; San Francisco (CA); On-site; USD 150,000 - 190,000; Full time

Job summary

Brahma Consulting Group is seeking an experienced Perception Engineer to own and scale our perception stack on NVIDIA edge hardware (Jetson Thor/Orin). You will ship real-time CV models, video ingestion frameworks, and tracking algorithms for field deployments.

This role emphasizes production ownership and cross-functional collaboration across hardware and software teams. The candidate should bring 4–7 years of hands-on experience in applied perception or ML systems, strong C++ and Python

Qualifications

• 4–7 years of hands-on experience building and deploying applied perception or machine learning systems in production at a startup or high-growth tech environment.
• Advanced proficiency in C++ (C++20 preferred) and Python.
• Direct experience deploying on edge hardware using NVIDIA DeepStream, GStreamer, TensorRT, and JetPack (Jetson Orin/Thor).
• Background in computer vision for hardware products in safety-critical or hardware-adjacent fields (e.g., robotics, autonomous vehicles, aerospace, or physical security).
• Strong knowledge of video streaming formats, camera feeds, multi-object tracking, and multi-camera association.

Responsibilities

• Perception Architecture: Re-architect and optimize core perception services running on NVIDIA Thor to support increased camera stream capacity, lower latency, and stability.
• Model Optimization & Deployment: Evaluate, swap, and deploy faster, higher-performing deep learning models for real-time edge performance.
• 3D Perception & Tracking: Implement monocular depth estimation on the NVIDIA Thor to enable robust 3D tracking across property boundaries.
• Pipeline Engineering: Build and maintain high-throughput video ingestion and media processing pipelines using DeepStream, GStreamer, and FFmpeg.
• Cross-Functional Execution: Collaborate with device, hardware, and systems teams to unblock dependencies and push updates to customer units.

Skills

C++

Python

NVIDIA edge stack

DeepStream

GStreamer

FFmpeg

TensorRT

Jetson Thor/Orin

3D perception

Multi-object tracking

Video pipelines

Tools

PyTorch

TensorRT

JetPack

GStreamer

FFmpeg

NVIDIA DeepStream

Job description

We are seeking an experienced, product-focused Perception Engineer (4–7 years of experience) to own, evolve, and scale the perception stack powering our autonomous home security hardware.

This is an applied, production-first engineering role. You won't be doing theoretical lab research—you will own the end-to-end perception pipeline and ship real-time computer vision models, video ingestion frameworks, and tracking algorithms directly onto NVIDIA edge hardware (Jetson Thor/Orin) deployed in the field.

Key Responsibilities

• Perception Architecture: Re-architect and optimize our core perception services running on NVIDIA Thor to support increased camera stream capacity, lower latency, and maximize pipeline stability.


• Model Optimization & Deployment: Evaluate, swap, and deploy faster, higher-performing deep learning models (e.g., PyTorch to TensorRT conversion) while maintaining real-time performance limits on edge hardware.


• 3D Perception & Tracking: Implement monocular depth estimation onto the NVIDIA Thor platform to enable robust 3D tracking of people and vehicles across property boundaries.


• Pipeline Engineering: Work deeply with NVIDIA DeepStream, GStreamer, and FFmpeg to build and maintain high-throughput video ingestion and media processing pipelines.


• Cross-Functional Execution: Collaborate directly with device, hardware, and systems engineering teams to unblock dependencies and push reliable software updates to customer units.

What We're Looking For

Requirements

• Experience: 4–7 years of hands-on experience building and deploying applied perception or machine learning systems in production at a Series A–D startup or high-growth tech environment.


• Core Languages: Advanced proficiency in C++ (C++20 preferred) and Python.


• NVIDIA Edge Stack: Direct experience deploying on edge hardware using NVIDIA DeepStream, GStreamer, TensorRT, and JetPack (ideally Jetson Orin or Thor).


• Domain Background: Background in computer vision for physical hardware products in safety-critical or hardware-adjacent fields (e.g., robotics, autonomous vehicles, aerospace, or physical security).


• Video & Media Pipelines: Strong knowledge of video streaming formats, camera feeds, multi-object tracking, and multi-camera association.


• Languages: C++20, Python


• Media & Video Processing: GStreamer, FFmpeg