inkling-hire

inkling-hire v1

A LoRA adapter for Inkling (Thinking Machines Lab) specialized for résumé screening support: calibrated fit scores, evidence citation, principled abstention, and name-invariance.

First model in the Navox Labs inkling-* series. Built to power navox-labs/inkling-hire — an open-source screening system with an always-on counterfactual bias audit, blind review by default, and human-in-the-loop by construction. Navox Labs is not affiliated with or endorsed by Thinking Machines Lab.

What it does

Given a job posting and a résumé (the score_v1 prompt in the repo), it emits one JSON verdict: a calibrated fit probability, up to 5 quoted evidence items, job-relevant gaps, or an explicit insufficient_information abstention for unreadable input. Wrong-field candidates get low scores, not abstentions — identity signals are not scoring features.

Evaluation (vs base Inkling, held-out synthetic val, efforts 0.3/0.8)

The counterfactual gap — mean absolute score change when only the candidate's name is swapped across demographic groups (Bertrand & Mullainathan 2004 audit names) — improved and is stable across effort. Fairness non-regression was a hard ship gate. Details, methodology, and two honest caveats (a mis-specified Brier criterion; rubric-mastery vs world-mastery): finetune/RESULTS.md.

Training

• Data: 100% synthetic — 5,468 examples generated by a deterministic, open-source engine (no real personal data, reproducible byte-for-byte: python -m finetune.build_training_data, hash 16c9ba62c5d0993e). Three task types: rubric-target scoring (demonstrated skills weighted over merely listed ones), abstention discipline (incl. anti-abstention for wrong-field candidates), and counterfactual consistency (name-swapped twins with byte-identical targets).
• LoRA rank 32, 1 epoch, lr 2e-4 linear, batch 64, trained on Tinker (~$34 at July-2026 rates).
• Val split holds out entire (family, seniority) cells to test generalization.

Intended use & limits

Decision support for human recruiters, embedded in a pipeline that enforces review — never automated hiring decisions. This is both the design and a condition of Inkling's Acceptable Use Policy (no automated employment decisions; deployer disclosure duties), which binds derivatives including this adapter. Trained and evaluated on synthetic English résumés across 10 job families; real-world résumé performance is not claimed — run your own evaluation (the repo ships the audit harness to do it, per-run and free). Base-model limitations (hallucination, training-data bias) may persist; the counterfactual audit measures, not eliminates, name bias.

Format & deployment

This repo ships the adapter in Tinker LoRA format (adapter_model.safetensors

• adapter_config.json, ~20GB — rank 32 across all layers incl. 256 experts). Converting to standard PEFT format requires the base model's weights on local disk — which your serving machine needs anyway — so convert on the box that serves it (one command, via tinker-cookbook):

from tinker_cookbook import weights
weights.build_lora_adapter(
    base_model="thinkingmachines/Inkling",   # already in your HF cache if you serve it
    adapter_path="./inkling-hire",           # this repo, downloaded
    output_path="./inkling-hire-peft",       # vLLM --lora-modules / SGLang --lora-paths
)

Then serve with vLLM/SGLang and point the inkling-hire pipeline's config.yaml at your endpoint — no candidate data leaves your infrastructure. (On Tinker, the checkpoint is directly usable at tinker://2b4f6b9a-9e2c-546c-a600-866e9755dbb8:train:0/sampler_weights/final.)

Citation & contact

• Author: Nahrin Oda (huggingface.co/nahrin)
• Research lineage: Winnow — open bias-audit harness for AI résumé screeners · The AI Recruitment Crisis (Navox Labs)
• Code & pipeline: github.com/navox-labs/inkling-hire
• Questions, deployments, custom fine-tunes: [email protected] · Navox Labs