AI Research Engineer - DL/DRL Driven HFT

DeepFin Research is a proprietary high-frequency trading (HFT) firm powered by cutting-edge Deep Learning (DL) and Deep Reinforcement Learning (DRL). We do not rely on traditional quant strategies; instead, we build autonomous, self-learning trading agents that operate at the microsecond level. You will be joining an elite team of Kaggle Grandmasters and former senior researchers from DeepMind and Nvidia to build innovative AI strategies in quantitative finance.

We are looking for an exceptional AI Research Engineer to bridge the gap between groundbreaking AI research and ultra-low latency production systems. In this role, you will be the architectural backbone of our research team. You will own the end-to-end RL stack, scaling cutting-edge deep reinforcement learning models across multi-GPU clusters and optimizing them down to the bare metal using CUDA.

• End-to-End RL Infrastructure: Architect, build, and maintain our proprietary end-to-end DRL training and inference pipelines.

• Distributed Training: Design and scale multi-GPU and multi-node training systems for massive parameter models, ensuring maximum throughput and minimal bottlenecks.

• Hardware-Level Optimization: Write, profile, and optimize custom CUDA kernels to accelerate both the training of complex neural networks and the ultra-low latency inference engine used in live trading.

• Research Collaboration: Work side-by-side with our senior researchers (ex-DeepMind, ex-Nvidia) and Kaggle Grandmasters to translate mathematical formulations and algorithmic prototypes into highly efficient, production-ready code.

• Environment & Simulation: Optimize our custom market-simulation environments, ensuring massive parallelization for RL agent rollouts and experience replay.

• Educational Background: MS or Ph.D. in Computer Science, Artificial Intelligence, Machine Learning, Computer Engineering, or a highly related field.

• Deep Learning & RL Expertise: Deep understanding of modern DL and DRL algorithms (PPO, SAC, DQN, Transformer architectures) and frameworks (PyTorch, JAX).

• High-Performance Computing: Proven track record with distributed training paradigms (DDP, FSDP, Megatron, Ray) and multi-GPU orchestration.

• Systems Programming: Exceptional skills in C++ and Python.

• CUDA Mastery: Hands-on experience writing custom CUDA kernels and profiling GPU performance (Nsight Systems, Nsight Compute).