Generative AI Developer, full-time and fully remote, with pay that goes up to $135,000 a year.
Someone in this role spends their days building the plumbing that connects large language models and image-generation systems to the actual products people use. Not research, not a paper to publish. Working software that a customer touches, day in and day out, whether that's a support chatbot, an image-generation feature inside a design tool, or something in between.
What you'd be building
The work centers on integrating generative model capabilities into real applications and ensuring they behave well once they're live. That's a different skill than getting a demo working in a notebook. A model that answers questions beautifully in testing can still fall apart once real users start typing things nobody predicted, feeding it edge cases and odd phrasing that never appeared in the dataset.
- Build and integrate generative model features directly into applications
- Fine-tune models for the specific use cases they're being deployed for
- Tune inference performance and manage cost once a feature is running in production
Say a chatbot feature starts confidently inventing product specs that don't exist. Fixing that isn't just a prompt tweak; it usually means rethinking how the model's outputs get grounded in real data before they ever reach a user. That kind of debugging is a normal week here, not an emergency.
Skills that matter for this seat
Python is the baseline language for almost everything on this team. Beyond that, comfort with large language models and diffusion models, solid prompt-engineering instincts, real API integration experience, working knowledge of at least one cloud platform, and hands-on fine-tuning experience all show up regularly in the day-to-day. None of these need to be equally deep. Someone might come in strong on the LLM side and lighter on diffusion models, or the other way around, and that's a normal starting point rather than a gap that rules someone out.
Two years of hands-on experience building with generative AI models is the minimum here, whether that's LLMs, image generation systems, or both. Naukri Mitra weights actual shipped work over academic exposure when narrowing down candidates, since much of what matters in this job only shows up once a model has to hold up under real traffic rather than in a controlled demo. On the education side, a bachelor's degree is expected, most often in computer science or a closely related field, though the exact major matters less than what a candidate has actually built with it. A self-taught developer with a strong portfolio of shipped generative AI work is a realistic fit here, provided the degree box still gets checked somewhere.
Pay and how the work is structured
This is full-time work, paying up to $135,000 annually, and it's remote with no location requirement. Model fine-tuning and API integration knowledge get used constantly here, not as a checkbox skill but because half the job is stitching third-party model APIs into systems that weren't originally designed around them. There's a fair bit of ambiguity in how a given week unfolds. Some weeks are mostly heads-down coding; others get eaten by a production model that started drifting and needs attention fast. Neither kind of week is more or less normal than the other; the mix is just part of the role.
A distributed team means the actual working hours are flexible within reason, as long as there's overlap with the rest of engineering for the moments that need real-time back-and-forth. Nobody's expected to be online at a specific hour just to prove they're working; what matters is whether the fine-tuning job finished, whether the integration is stable, and whether a drifting model got caught before it caused a real problem downstream.
What comes with it
Full-time here means the standard package plus a couple of things specifically aimed at this kind of work.
- Health insurance
- Paid time off
- Retirement plan matching
- Genuine remote-work flexibility around when hours get logged
- A stipend that covers compute costs or attending an AI conference
The compute stipend, in particular, matters more here than it would in a typical software role, since fine-tuning work and inference testing can rack up real cloud costs quickly if nobody's covering them. A single week of aggressive fine-tuning experiments on a mid-sized model can easily run into a few hundred dollars of compute time, and that cost shouldn't come out of anyone's own pocket just to do the job properly.
Who you'd be working alongside
This isn't a huge team, and that's mostly a good thing. Decisions about which model to use, how aggressively to fine-tune it, or when a feature is ready to ship don't have to move through five layers of sign-off. It does mean less hand-holding than a bigger company might offer, so someone who wants a fully mapped-out onboarding plan for their first month might find the ramp-up here a little looser than expected. Most new hires end up learning the internal systems by working on a real feature within the first couple of weeks rather than sitting through a long training track first.
Generative AI moves fast enough that whatever's considered best practice in prompt design or fine-tuning today can look dated within a couple of quarters. People who do well in roles like this one are usually the ones who treat churn as normal rather than frustrating and don't need a settled, static tech stack to feel productive. A framework or library that's central to the workflow this year might get swapped out next year, and that's just the shape of the field right now, not a sign that anything here is unstable.
Applying
Send a resume along with a link to something you've actually shipped: a project, a repo, a product feature, anything that used a generative model in a real way rather than a tutorial walkthrough. If you've fine-tuned a model and can talk through what changed and why, that goes a long way here. There's no fixed application deadline attached to this posting. Reviews happen on a rolling basis, and an earlier application is generally reviewed sooner than one that comes in weeks later.
