Neuronify

Fine-tuning

Continuing to train an existing model on your own examples so it adopts a format, tone, or narrow skill that prompting alone does not reliably produce.

Hiring data as of September 20, 2026 · refreshed daily

companies naming it
12 of 47companies naming it
open roles naming it
22open roles naming it
of those based in NYC
10of those based in NYC
pay: under 8 posted ranges
—pay: under 8 posted ranges

What Fine-tuning means

Fine-tuning takes a pretrained model and trains it further on a smaller, task-specific dataset. It changes the model’s weights, unlike prompting or retrieval, which only change its input. Parameter-efficient methods such as LoRA make it affordable by training a small set of added weights instead of the whole model.

It is the right tool for consistent style and format, for narrow classification or extraction tasks at high volume, and for making a small model do what otherwise needs a large one. It is the wrong tool for giving a model new facts — that is what retrieval is for — and it carries a maintenance cost, because the tuned model must be redone when the base model is replaced.

Fine-tuning in New York job posts

12 of the 47 New York AI companies we track name Fine-tuning in at least one open role today, across 22 postings. 10 of those are based in New York; the rest are remote or in the companies’ other offices.

This is engineering vocabulary: 64% of the postings that name it are Engineering roles.

How to read it

We count fine-tuning only where the posting places it next to a model, since sales postings “fine-tune” messaging. A genuine mention points to a team with its own training data and enough inference volume that a smaller tuned model pays for itself. It is a meaningful step up in ML depth from application-only roles.

What these roles pay

Not enough data to publish. We need at least eight distinct posted salary ranges on New York roles naming Fine-tuning before showing a median, and today there are fewer. For posted pay by function and seniority across the whole market, use the NYC salary calculator or the NYC AI engineer salary guide.

Trend

We began extracting this term from posting text on September 20, 2026. Job descriptions disappear when a role is filled, so the series cannot be backfilled; a trend line appears here once a week of data exists.

Who names it

Open roles naming Fine-tuning, by company.

  • AlphaSense6
  • Normal Computing3
  • Hume AI2
  • Mirage (fka Captions)2
  • Modal2
  • Comet1
  • EvolutionIQ1
  • Headway1
  • K Health1
  • Maven Clinic1
  • Ramp1
  • Runway1

Which roles

The same postings, by function and by seniority.

  • Engineering14
  • Research4
  • Other2
  • GTM1
  • Product1
  • Mid16
  • Staff/Principal4
  • Senior2

Named alongside Fine-tuning

Terms that appear in the same postings far more often than chance would put them there. The number is how many of the 12 companies pair the two.

Open roles naming Fine-tuning

A sample from today’s data, New York roles first and one per company before any repeats. Links go to the employer’s own posting. For the full market, see AI jobs in NYC.

How this is counted

Every day we read the open roles on the public job boards of the New York AI companies in our coverage — the NYC AI 100 and the AI in NYC Show roster. This term is matched against each role’s description after removing the text a company repeats across its postings, so an “About us” paragraph cannot tag every role the company has open. A vendor’s own postings never count toward its own name.

The figure means named in a job posting. It does not mean used in production, and a “nice to have” counts the same as a requirement. Companies are the headline number because posting counts are dominated by whichever few employers are hiring hardest this month.

Full method, including the pay rules, is on the glossary index; the underlying series is the NYC AI Hiring Index.

Fine-tuning — common questions

What does Fine-tuning mean in an AI job posting?

Continuing to train an existing model on your own examples so it adopts a format, tone, or narrow skill that prompting alone does not reliably produce. We count fine-tuning only where the posting places it next to a model, since sales postings “fine-tune” messaging. A genuine mention points to a team with its own training data and enough inference volume that a smaller tuned model pays for itself. It is a meaningful step up in ML depth from application-only roles.

How many New York AI companies are hiring for Fine-tuning?

As of September 20, 2026, 12 of the 47 New York AI companies we track name Fine-tuning in the description of at least one open role, across 22 postings (10 based in New York). The count refreshes daily from the companies’ own job boards.

What do roles that ask for Fine-tuning pay in New York?

We do not publish a pay figure for Fine-tuning yet. Fewer than eight New York postings naming it carry a distinct posted salary range, and below that number a median says more about two or three employers than about the market.

Which skills are asked for alongside Fine-tuning?

In the same postings, the terms most distinctively paired with Fine-tuning are Reinforcement learning, LangChain, Multimodal AI, MLOps, RAG. We rank pairings by how much more often they appear together than apart, so near-universal terms such as Python do not crowd out the informative ones.

Need someone who has actually shipped this?

A keyword in a posting is easy to match and hard to verify. We match companies with AI engineers and consultancies vetted on production work, and tell you when the project needs a different skill than the one you named.

Start a project briefJoin the expert network