AI Engineer
We are pioneering an AI-native organization, building AI native services architectures grounded in the heart of real-world commerce. Operating at the intersection of frontier tech and market dynamics, our systems continuously ingest, process, and iterate on high-density commercial data streams in real time.
Responsibilities & requirements
Key responsibilities
- Post-Training & RL Optimization: Architect, fine-tune, and align foundation models using advanced Post-Training pipelines, Reinforcement Learning (RLHF/RLAIF, DPO, PPO), and reasoning-focused workflows.
- End-to-End Commercial Deployment: Own the entire lifecycle of core AI systems—from algorithmic formulation and data synthesis to high-performance inference and production deployment.
- Research-to-Product Translation: Rapidly absorb and implement cutting-edge papers and research breakthroughs, adapting them into scalable, robust commercial solutions.
- Algorithmic Problem Solving: Tackle hard, open-ended technical bottlenecks independently, continually raising the bar for model performance, precision, and latency.
Requirements
Technical track record (must-have)
- Expertise in frontier AI: Deep hands-on experience in Post-Training (SFT, Preference Alignment) and Reinforcement Learning (RL algorithms, reward modeling, search/reasoning frameworks).
- Proven commercial delivery: A solid history of successfully shipping end-to-end commercial AI products or high-impact production systems (not just offline demos or benchmark toy models).
- Hands-on mastery: Exceptional coding and engineering fundamentals (Python, PyTorch, CUDA, distributed training frameworks like Ray, DeepSpeed, or Megatron-LM).
Research & mindset traits
- First-principles research depth: Possess a “research-grade” understanding of AI fundamentals. You know why a model behaves the way it does under the hood, not just how to call a framework.
- Scholarly credentials: Published papers in top-tier AI/ML venues (e.g., NeurIPS, ICML, ICLR, CVPR, ACL) or an equivalent track record of novel algorithmic implementations.
- Prototyping to production: True hands-on grit. You love getting your hands dirty with raw code, noisy data, and custom GPU profiling.
- Elite “solo-solver” ability: Exceptional autonomous problem-solving skills. You can be dropped into an ambiguous, uncharted problem and independently figure out the solution from zero to one.