Parse a Resume to JSON in Node.js

To parse a resume in Node.js, send the file as multipart form data to the SharpAPI parse_resume endpoint, then poll the returned status URL until the job reports success. Native fetch handles both calls; the result is 50+ structured candidate fields.

The whole flow, native fetch

import { readFile } from 'node:fs/promises';

const API_KEY = process.env.SHARP_API_KEY;
const headers = { Authorization: `Bearer ${API_KEY}` };

// 1. Submit — 202 Accepted + status URL, immediately
const form = new FormData();
form.append('file', new Blob([await readFile('resumes/candidate.pdf')]), 'candidate.pdf');
form.append('language', 'English');

const submit = await fetch('https://sharpapi.com/api/v1/hr/parse_resume', {
    method: 'POST',
    headers,
    body: form,
});
const { status_url } = await submit.json();

// 2. Poll — jobs typically finish in seconds
let job;
do {
    await new Promise((resolve) => setTimeout(resolve, 2000));
    job = await (await fetch(status_url, { headers })).json();
} while (job.data.attributes.status === 'pending');

const candidate = job.data.attributes.result;
console.log(candidate.candidate_name);
candidate.positions.forEach((role) =>
    console.log(`- ${role.position_name} @ ${role.company_name}`),
);

Or take the npm package

@sharpapi/sharpapi-node-parse-resume (built on @sharpapi/sharpapi-node-core) wraps the same flow with rate limiting and typed DTOs — worth it once resume parsing stops being a one-off script.

Typing the response

The schema is deterministic — every parse returns the same structure, with empty values where the document lacks data. That makes TypeScript pleasant: one interface, written once against the sample payload, describes every response your pipeline will ever see. Start from the field reference.

Production notes

• Webhooks over polling. Register a webhook URL at submission and delete the do-while loop — one Express route receives every result. Mandatory reading before a backfill: bulk resume parsing.
• Scans included. JPG, PNG and TIFF files go through the identical call — OCR runs server-side.
• Errors are statuses. A corrupt or password-protected file resolves to a failed status, not a hang; branch on it and keep the queue moving.