Build the experimental infrastructure behind the research.
Latent Minds Institute combines an open research programme on interpretability, model cognition and safety with a product for building and investigating open models.
Research
Senior Research Engineer
Own the experimental infrastructure behind the research programme: Crucible environments, evaluation batteries and the pipelines that turn behavioural runs into published, checkable evidence.
What you would build
- New Crucible game environments and their run harnesses
- Probing and steering pipelines on open models, with causal checks
- Red teaming and refusal consistency suites
- Evidence releases: raw records, summaries, reproduction paths
Evidence of ability
- One experimental harness or evaluation suite you built and can walk through
- Writing that separates behavioural from mechanistic claims
Product
Member of Technical Staff · Product
Build the product: the guided environment that takes a team from an open checkpoint to an evaluated, investigated, releasable model, with every stage attached to one record.
What you would build
- The model record: lineage, runs, evaluations, releases
- Training orchestration over LLaMA Factory and rented or local runners
- Evaluation matrices, prompt level diffs and release gates
- Export paths: weights, vLLM configs, evidence bundles
Evidence of ability
- Product engineering you shipped and still stand behind
- Fluency with the open model stack you would be orchestrating
- Experience designing interfaces that surface uncertainty, provenance and limitations accurately
If neither fits but your work belongs here, send a general application below with one thing you have built.