Founding AI Backend Engineer

Pakket•San Francisco, CA

About The Position

Pakket is building the operating system for the physical world of warehouses. Today, most warehouse operations run on a patchwork of spreadsheets, radios, and outdated software. The actual activity on the floor — forklifts moving, trucks arriving, people working — is largely invisible to software systems. Pakket uses cameras and AI to create a real-time, live map of warehouse operations. They track forklifts, people, trailers, and yard movement — turning physical activity into structured, actionable data. This unlocks real-time visibility into operations, automated alerts (delays, congestion, detention risk), historical analytics on performance, and a system of record for the warehouse floor. The goal is to make every warehouse measurable, optimizable, and autonomous. Pakket is already deployed in warehouses across America and is scaling fast.

Requirements

3–6 years of backend or full-stack engineering experience
Strong Python — you write clean, readable code and can navigate an existing codebase quickly
Experience with FastAPI or similar async Python frameworks, and SQL databases
Solid Linux skills — comfortable SSHing into a remote machine, debugging processes, and managing services
Docker proficiency — you containerize applications and understand multi-service setups
Experience with cloud infrastructure (AWS, GCP, Render, DigitalOcean, or similar)
A debugging mindset — when something breaks at 2am in a warehouse, you figure it out
Comfortable with ambiguity and building in an early-stage environment

Nice To Haves

Worked with computer vision libraries — OpenCV, YOLO/Ultralytics, or similar
Experience with GPU compute, CUDA, or TensorRT
Worked with RTSP video streams, FFmpeg, or HLS pipelines
Deployed and maintained systems on edge hardware outside of a data center
Experience with real-time systems — WebSockets, pub/sub, or streaming architectures

Responsibilities

Debug and improve the computer vision pipeline — including YOLO-based detection and geometric positioning
Own backend reliability — FastAPI, PostgreSQL, async SQLAlchemy, and the edge-to-cloud data pipeline
Build and ship customer-facing features: alerts, dashboards, historical playback, and zone monitoring
Manage edge server deployments at warehouse sites — Linux configuration, process management, remote access
Containerize and harden the system using Docker for consistent deployments across edge hardware
Identify and fix performance bottlenecks across compute, VRAM, storage, and networking
Work with the founder on architecture decisions as the system scales from 1 warehouse to many