Focuz is a well-funded, early-stage startup on a mission to redefine video intelligence. We are building a next-generation platform that transforms raw video streams from any camera into structured, actionable insights. Moving beyond simple object detection, we are creating a system that understands context, behavior, and patterns, enabling businesses and consumers to make smarter decisions. We are a small, agile team of builders and innovators, and we're looking for foundational members to help us shape the future of video technology.
We're looking for an Applied LLM Python Engineer to join our team at Focuz. We need specialist who can ship production code that leverages LLMs to analyze video footage in real-time, so it's not a research position. You'll be building systems that allow users to query security cameras using natural language and get intelligent insights about what's happening in their spaces.
What You'll Be Doing:
- Writing Python code that orchestrates multiple LLMs to analyze video clips (15 seconds to 15 minutes)
- Building context injection systems that add business rules, schedules, and location-specific information to prompts
- Creating evaluation frameworks to measure if our LLMs are correctly identifying safety violations, suspicious behavior, or operational issues
- Optimizing costs by choosing the right model for each task (when to use GPT-4 vs Claude Haiku vs Llama)
- Building robust production systems that handle API failures, rate limits, and inconsistent model outputs
Requirements:
- Strong Python engineering experience: ability to write clean, maintainable code for production-level applications, not just experimental notebooks.
- Production LLM experience: track record of shipping features with LLM APIs from major providers like OpenAI, Anthropic, and Google.
- Multi-modal expertise: understanding of how to work with models such as GPT-4V, including feeding them images and video frames.
- Context engineering skills: capability to structure large, complex prompts (50KB+) with effective examples, rules, and context.
- Evaluation mindset: A focus on metrics and the ability to build systems to measure and optimize LLM performance.
Bonus points:
- Experience with video processing (OpenCV, FFmpeg);
- Knowledge of open-source LLMs and local deployment;
- Understanding of structured output generation (JSON mode, function calling);
- Experience with async Python and streaming responses.
#focuz
Signaler l'offre
