Research Scientist/ML Scientist (LLM/RL/Post-Training/DPO/GRPO) - Up to $400K + equity - Remote

Attis; United States; Remote; USD 250,000 - 400,000; Full time

Benefits offered by this job

Equity

Remote in USA

Competitive salary

Job summary

Attis is seeking a Research Scientist (LLM/RL/Post-Training/DPO/GRPO) to advance edge AI at scale, remotely from the USA. You will work on open-weight model post-training, alignment, and model compression across end‑to‑end pipelines, collaborating with engineers and researchers to push state-of-the-art capabilities.

Ideal candidates have hands-on experience with distributed GPU clusters, PyTorch loss‑function optimization, and KV‑cache tuning for fast inference.

Qualifications

• Hands-on experience fine-tuning open-weight models using alignment techniques (GRPO, DPO, RLVR) or model compression methods.
• Experience training models across distributed GPU clusters (16+ H100/B200) with Slurm, DeepSpeed, or FSDP.
• Proven ability to optimize serving, rollouts and KV-cache with vLLM, SGLang, or TensorRT‑LLM.
• Preference for candidates with clearance related to US Space Force/DoD programs.

Skills

LLMs

Reinforcement Learning

RLHF

Post-Training

Model Compression

Quantization

Distillation

Distributed Training

Edge AI

Tools

Slurm

DeepSpeed

FSDP

vLLM

TensorRT-LLM

Job description

Research Scientist (LLM/RL/Post-Training/DPO/GRPO), Remote, USA (Between $250-400K DOE + equity)

Research Scientists sought for a well‑funded, highly successful defense AI scaleup who bring state of the art AI to the tactical edge. They are a talented team of engineers and researchers who perform the full stack of model training to improve models every day. We are seeking true Model Builders - scientists who operate at the PyTorch loss-function and kernel layer to execute the full stack of open-weight LLM post‑training, alignment, and model compression every single day.

The company is developing advanced AI systems that can operate directly at the edge, bringing high‑performance AI into demanding real‑world environments without relying on constant cloud connectivity. Its technology enables sophisticated models to run securely across laptops, embedded devices, robotic platforms and other hardware, with models optimized for specific applications and operational requirements.

Offering a base of between 250-400K depending on experience + generous equity, fully remote working in the US and other benefits.

Ideally you will have experience performing experimentation in one or more of the following areas:

• Post-Training & Alignment: Hands‑on experience adapting or fine‑tuning open‑weight models (Qwen, Llama, DeepSeek) using modern alignment techniques (GRPO, DPO, RLVR) OR model compression methods (Quantization, Distillation, FP8/AWQ) instead.
• Cluster Scale Compute: Experience training models across distributed GPU clusters (16+ H100/B200 nodes) using Slurm, FSDP, or DeepSpeed.
• Rollout Mechanics: Expertise in serving, parallel worker rollouts, and KV‑cache optimization via vLLM, SGLang, or TensorRT‑LLM.
• Domain Alignment (Preferred but NOT essential): Prior exposure to mission‑critical applications across Robotics, Tactical Edge AI, Autonomous Systems, or Defense (US Space Force, DoW, DoD, DARPA, JHU APL, Sandia).

We have a strong preference for candidates with an active clearance to work on projects related to US Space Force, DoW, DoD etc.

Key Skills

Large Language Models, LLM, Foundation Models, Generative AI, Reinforcement Learning, RLHF, Reinforcement Learning from Human Feedback, Post-Training, Pretraining, Midtraining, Fine‑Tuning, Model Adaptation, Model Training, Distributed Training, Multimodal, Vision-Language Models, VLM, Vision Model Training, Machine Translation, Multilingual Models, Model Evaluation, LLM Evaluation, Benchmarking, Benchmark Development, Data Ablation, Model Distillation, Quantization, Model Compression, Robotics + LLM, Edge AI

Disclaimer:

Attis Global Ltd is an equal opportunities employer. No terminology in this advert is intended to discriminate on any of the grounds protected by law, and all qualified applicants will receive consideration for employment without regard to age, sex, race, national origin, religion or belief, disability, pregnancy and maternity, marital status, political affiliation, socio‑economic status, sexual orientation, gender, gender identity and expression, and/or gender reassignment. M/F/D/V. We operate as a staffing agency and employment business. More information can be found at attisglobal.com.