AI Engineer
Posted Updated
Description
- Advertising is being rebuilt for a world where AI agents plan, buy, and create the ads, and InMobi is building for that world now
- As an AI Engineer, you’ll build the harnesses, agents, and fine-tuned models that bring generative AI into our products and internal workflows
- You’ll work across the full stack of modern GenAI systems: designing agent harnesses, fine-tuning LLMs and diffusion models with RL, building evaluations, and optimizing systems both with and without training
- This isn’t a roadmap item
- AI buying agents already purchase our inventory through AdCP – the Ad Context Protocol
- Our own agents build complete campaigns, creatives included
- And inside the company, agents write, review, and ship code, build their own tools, and run customer success, ad quality, and cloud cost management
- We’ve built the foundation to push further, with GPU capacity to train and serve our own models and an AI gateway that handles model traffic for every team
- If you want to own hard problems from first idea to production, there’s no shortage of them here!
- Location: This role is based on-site at our San Mateo, CA office location
- Autonomous creative generation: Build a system that turns brand guidelines into ad creatives, then keeps improving them from live campaign performance
- ML research agent: Build an agent that speeds up ML research for our Applied Scientists
- Learning from real-world signals: Turn signals such as creative performance and developer feedback into training data, so our systems keep getting better after launch
- End-to-end ownership: Own projects from framing the problem and defining success through shipping and iterating in production
- Evals and measurement: Define what “good” looks like for systems without an obvious metric, and build the evals and annotation processes to measure it
- Open-source contribution: Build on and contribute to the open-source tools shaping this space, including training frameworks like LLaMA-Factory and verl and agent harnesses like opencode and OpenResearch
- Cross-team partnership: Partner with Applied Scientists, Product, and Engineering to turn promising prototypes into production systems
- A solid grasp of LLM and diffusion model architectures
- 2+ years of experience building GenAI systems in production
- Expertise in building agents and agent harnesses
- Experience building knowledge retrieval systems, including ANN-based and knowledge-graph-based retrieval
- Experience with no-train optimization techniques and frameworks like DSPy
- A Bachelor’s or Master’s degree in Computer Science, Machine Learning, or a related field, or equivalent practical experience
- Expertise in Python
- Hands-on experience fine-tuning LLMs or diffusion models with RL techniques (for example, PPO, GRPO) or preference optimization methods like DPO
- A track record of designing evals from scratch, including writing labeling guidelines and working with annotators
- Understanding of the ads ecosystem (OpenRTB, SSPs, DSPs, ad exchanges, MMPs)
- Contributions to open-source GenAI projects, such as LLaMA-Factory, verl, opencode, or OpenResearch
- Public work in LLMs, diffusion models, reinforcement learning, or agents, such as publications, write-ups, or talks
- Experience with large-scale GPU training and serving, including distributed training and inference optimization
- Experience building coding agents or developer tools that write, review, or ship code
Skills
Sign in to read and join the discussion.