We're hiring a Remote NLP Engineer, full-time, fully remote, $132,000 a year, open to candidates anywhere.
What you'll need
You'll need a bachelor's degree, usually in computer science, computational linguistics, or something in that neighborhood. Alongside that, we're looking for 2 years of experience building text classification, extraction, or generation systems. Proficiency with common NLP libraries and frameworks is expected coming in, since there won't be a long ramp-up period to get comfortable with the tooling.
Many strong candidates come from slightly different backgrounds. Some have spent their two years specifically in NLP. Others came from broader ML roles and picked up text-focused work along the way. Both are fine as long as the hands-on NLP experience is genuine.
The two years don't need to come from a single employer either. Someone who spent a year building a classification system at one company and another year on a generation project somewhere else brings a genuinely broader NLP background than someone who spent the same time on one narrow task repeatedly, and both get evaluated on the depth of the work rather than how neatly it fits on a resume.
Skills
- Python
- Natural language processing
- Transformers
- spaCy or NLTK
- Machine learning
- Text preprocessing
- Model evaluation
Text preprocessing sounds simple until you've been burned by it once. A sentiment classifier on one project scored well in testing but kept marking sarcastic complaints as positive after launch, because the training data barely contained any sarcastic examples, and the preprocessing pipeline had also been stripping out some of the punctuation cues that might have hinted at tone. Fixing the model architecture wouldn't have solved that. Fixing the data and the preprocessing did.
Transformers show up in most projects now, but that doesn't mean fine-tuning a huge model from scratch every time. Plenty of tasks here are handled well by a smaller, fine-tuned model that runs cheaper and faster, and knowing when the bigger model actually earns its cost, rather than defaulting to it out of habit, is part of doing this job well.
What you'll do
- Build and fine-tune models for text classification, sentiment analysis, and language generation
- Preprocess and evaluate text datasets before a model ever gets trained on them
- Integrate NLP capabilities directly into applications, not just leave them running in a research environment
Evaluation gets more attention here than a quick accuracy number might suggest. A model can score well on an aggregate metric while quietly failing on a specific category of input, like sarcasm, informal slang, or a particular language variant, and catching that kind of gap before launch matters more than chasing another decimal point on the headline number.
Fine-tuning work usually happens in small, iterative rounds rather than a single large training run. A dataset gets cleaned up based on what the last round's errors revealed, the model gets retrained, and the cycle repeats until performance on the categories that actually matter stops improving much further.
Pay and benefits
This role pays $132,000 a year. There's retirement plan matching, time off, and health coverage, plus a budget for courses or conferences to keep up with a fast-moving field. Remote-work flexibility comes standard rather than as something negotiated case by case.
Naukri Mitra is coordinating the hiring for this opening, and the course and conference budget has previously been allocated to specialized transformer fine-tuning workshops and industry NLP conferences.
If you've looked at NLP engineer remote salary numbers at the two-year mark, this offer sits comfortably above the median, reflecting how tight the market's gotten for engineers who can take NLP work from a research notebook into something reliably running in production.
How the work happens
This is a small NLP team embedded inside a larger product group, so you'll work directly with people outside the ML world more than you might expect. A support team might flag that a classifier's misreading a specific type of ticket, and figuring out whether that's a data problem, a preprocessing issue, or a genuine model limitation is a normal part of the investigation.
Language and text data come with their own mess that a lot of other ML work doesn't deal with as directly: typos, slang, code-switching between languages, and domain-specific jargon that a general-purpose model was never trained on. Handling that mess well, rather than assuming clean, well-formed input, is a key part of what makes this role different from many other ML engineering positions.
Among NLP engineer jobs worldwide, some stay closer to pure research, and others lean fully into production integration. This one sits closer to the production side, so if what excites you most is shipping something users actually interact with, rather than publishing benchmark results, it's a good match.
Applying
Send a resume along with a short account of an NLP system you built in which the initial results looked strong but revealed a blind spot you had to fix. Interviews include a technical round on NLP fundamentals and model evaluation, followed by a shorter conversation about integrating models into a real product.
