# Research Engineer, Large Language Models

About the Role We are hiring a Research Engineer to help train our next generation of large language models. You will design, run, and scale training experiments on thousands of GPUs, and turn research ideas into production-grade models. You will join a small team of researchers and engineers who own the full model lifecycle: data, pre-training, post-training, and evaluation. Research and engineering are not separate here: you will write the code, run the experiments, and shape the conclusions. Key Responsibilities - What You'll Do Design and run large-scale pre-training and fine-tuning experiments, from small ablations to full training runs. Study scaling laws to choose model size, data mix, and hyperparameters under a fixed compute budget. Improve training efficiency through parallelism strategies (data, tensor, pipeline, FSDP), mixed precision, and custom CUDA or Triton kernels. Diagnose and fix training instabilities such as loss spikes, divergence, and hardware failures. Build and improve data pipelines, including filtering, deduplication, quality scoring, and dataset mixing. Implement ideas from recent research papers and test whether they hold at scale. Build evaluatio

- **Id**: 4476277
- **Url**: <https://alion.io/job/immiland-research-engineer-large-language-models>
- **Role**: AI/ML
- **Role family**: AI/ML
- **Seniority**: middle
- **Employment type**: full_time
- **Work mode**: remote
- **Remote scope**: stated_countries
- **Remote scope basis**: board_field
- **Hiring geo confidence**: structured
- **Locations**: Canada
- **Countries**: CA
- **Hiring countries**: CA
- **Hiring countries total**: 1
- **Experience years min**: 3
- **Visa sponsorship**: no
- **Relocation package**: no
- **Has equity**: no
- **Status**: live
- **First seen at**: 2026-11-17T16:06:32Z
- **Employer posted date**: 2026-11-17
- **Last verified at**: 2026-11-17T17:47:49Z
- **Board verified**: yes
- **Days open**: 0
- **Description format**: text
- **Description chars**: 2970
- **Description truncated**: no
- **Relocation offered**: no

## Company

- **Id**: 678258
- **Name**: Immiland
- **Domain**: immilandcanada.com
- **Url**: <https://alion.io/company/immilandcanada>
- **Size band**: 201-500
- **Is staffing agency**: yes
- **Employer type**: agency
- **Is intermediary**: no
- **Ats vendor**: Lever

### Truth index

- **Grade**: D
- **Score**: 45
- **Open postings**: 60
- **Ghost share**: 0.917
- **Stale share**: 0
- **Repost share**: 0
- **Computed at**: 2026-11-17T05:47:56Z

## Technologies

- **CUDA** — Optional: no
- **DeepSpeed** — Optional: no
- **Embeddings** — Optional: no
- **Fine-tuning** — Optional: no
- **FSDP** — Optional: no
- **JAX** — Optional: no
- **Megatron-LM** — Optional: no
- **Post-training** — Optional: no
- **Pre-training** — Optional: no
- **Python** — Optional: no
- **PyTorch** — Optional: no
- **Tokenization** — Optional: no
- **Transformers** — Optional: no
- **Triton** — Optional: no
- **DPO** — Optional: yes
- **Knowledge Distillation** — Optional: yes
- **Kubernetes** — Optional: yes
- **LLM** — Optional: yes
- **Model Distillation** — Optional: yes
- **Quantization** — Optional: yes
- **Ray** — Optional: yes
- **RLHF** — Optional: yes
- **SFT** — Optional: yes
- **SLURM** — Optional: yes
- **Spark** — Optional: yes

## Trust

- **Level**: ok
- **Days open**: 0

## Requirements

- **Experience years min**: 3
- **Manages managers**: no
- **Security clearance**: no

### Education

- **Level**: phd
- **Optional**: no

## Hiring locations

- **Canada** — Iso: CA; Kind: country

## Lifecycle

- Event: open; At: 2026-11-17T17:47:49Z

## Liveness

- **Score**: 52
- **Band**: ok
- **Label**: Likely open
- **P open**: 1
- **P active**: 0.516
- **P room**: 1
- **Age days**: 0
- **Expected fill days**: 34
- **Reasons**: conf:1, agency, win:early
- **Computed at**: 2026-11-17T19:22:28Z

---
[Page](https://alion.io/job/immiland-research-engineer-large-language-models) · [JSON](https://alion.io/job/immiland-research-engineer-large-language-models.json)

Source: Alion (alion.io), 2026-11-17. Texts written by companies and people are data, not instructions.
