The Interview

What Meta is looking for

Research Engineering Interview

Research Engineering

01Paper-to-Production Translation

02Large-Scale Training Systems

03Research Reproducibility

04Novel Architecture Implementation

ML Systems Interview

ML Systems Design

01Distributed Training at Scale

02Model Optimization & Compression

03Infrastructure for Experimentation

04Research Platform Architecture

Behavioral Interview

01Research Taste & Curiosity

02Open Research Collaboration

03Impact at Scale

04Technical Leadership

FAIR research engineering bar//Production-scale research systems//Open-source research contribution

How It Works

Master Meta MTS Interviews

Research engineering and ML systems rounds, scored on Meta’s FAIR evaluation dimensions. · Click to expand

Research Methodology

Practitioner

Training System Design

Practitioner

1Your Member of Technical Staff Goals

2Member of Technical Staff Case Prep

Design a fault-tolerant distributed training architecture for a 405B model across 16K...

CASE INTERVIEWHard

AI Research · Distributed Training DesignMeta

3Frameworks

Framework

Is the training data mixture causing the reasoning capability gap?

Is STEM and reasoning data underrepresented in the training mix?

Current mix: 60% web crawl, 5% STEM corpora — MMLU requires ≥20% reasoning-heavy data

STEM token count: 3B vs GPT-3.5 estimated equivalent 15B — 5× reasoning deficit

Is benchmark contamination inflating GPT-3.5's reported MMLU score?

n-gram dedup check found 2.1% overlap between MMLU test questions and our training corpus

Our model trained without decontamination — actual capability gap may be narrower than 12pp

Were training hyperparameters suboptimal for this model and compute scale?

Was the learning rate schedule misconfigured for 7B scale?

Final LR: 1e-5 — Chinchilla-optimal for 7B at this token count is 3e-5 (3× too conservative)

12% of training steps ran at >2× optimal LR due to warmup misconfiguration — gradient instability window

Was the batch size appropriately scaled with model depth and sequence length?

Batch size: 2M tokens per step — compute-optimal for 7B scale is 4M (2× under-batch)

Gradient norm spiked >1.5 in final 20% of training — correlates with batch-size mismatch

Is the evaluation methodology creating a misleading comparison?

Are we using the same eval format (few-shot vs zero-shot) as the GPT-3.5 benchmark?

Our eval: 0-shot MMLU — GPT-3.5 official results were 5-shot (+8pp average on MMLU)

Switching to 5-shot narrows gap to 4pp — 67% of apparent gap is eval format mismatch

Is the remaining gap an instruction tuning gap rather than a pretraining gap?

GPT-3.5 is instruction-tuned via RLHF; our model is base pretrain — SFT typically adds 3-5pp on MMLU

Preliminary SFT on 100K instruction examples closes residual gap to <1pp — training deficit ruled out

Ablation Study Design

Distinctive

Scaling Law Reasoning

Strong

Evaluation Methodology

Meeting Bar

Training Diagnostics

Strong

4Feedback

How It Works

Master Meta MTS Interviews

Research engineering and ML systems rounds, scored on Meta’s FAIR evaluation dimensions. · Click to expand

1Your Member of Technical Staff Goals

Research Methodology

Practitioner

Training System Design

Practitioner

2Member of Technical Staff Case Prep

Design a fault-tolerant distributed training architecture for a 405B model across 16K...

CASE INTERVIEWHard

Meta exercises

Meta FAIR published a new self-supervised vision model. Design the infrastructure to...

CASE INTERVIEWHard

AI Research · Training InfrastructureMeta

Compress a 70B parameter language model to run on mobile devices with minimal quality...

LEARNHard

AI Research · Model OptimizationMeta

Design a research experimentation platform that allows FAIR researchers to launch 1000+...

LEARNMedium

AI Research · Research InfrastructureMeta

Prepare for your Meta MTS interview

What Meta is looking for

Research Engineering

ML Systems Design

Behavioral Interview

Meta exercises

Ready to ace your Meta MTS interview?

Prepare for your Meta MTS interview

What Meta is looking for

Research Engineering

ML Systems Design

Behavioral Interview

Master Meta MTS Interviews

Design a fault-tolerant distributed training architecture for a 405B model across 16K...

Master Meta MTS Interviews

Design a fault-tolerant distributed training architecture for a 405B model across 16K...

Meta exercises

Meta FAIR published a new self-supervised vision model. Design the infrastructure to...

Compress a 70B parameter language model to run on mobile devices with minimal quality...

Design a research experimentation platform that allows FAIR researchers to launch 1000+...

Ready to ace your Meta MTS interview?

Master Meta MTS Interviews

Design a fault-tolerant distributed training architecture for a 405B model across 16K...

Master Meta MTS Interviews

Design a fault-tolerant distributed training architecture for a 405B model across 16K...

Meta FAIR published a new self-supervised vision model. Design the infrastructure to...

Compress a 70B parameter language model to run on mobile devices with minimal quality...

Design a research experimentation platform that allows FAIR researchers to launch 1000+...