diff --git a/backend/pdf-utils.js b/backend/pdf-utils.js
new file mode 100644
index 0000000..7c2a932
--- /dev/null
+++ b/backend/pdf-utils.js
@@ -0,0 +1,134 @@
+/**
+ * pdf-utils.js — Thin wrappers around poppler-utils CLI tools.
+ *
+ * Uses pdfinfo + pdftoppm (from poppler-utils) for PDF metadata extraction
+ * and page rendering. No external Node dependencies required.
+ */
+
+const { execFile } = require('child_process');
+const { promisify } = require('util');
+const fs = require('fs');
+const os = require('os');
+const path = require('path');
+
+const execFileAsync = promisify(execFile);
+
+const TIMEOUT_MS = 60_000;
+
+/**
+ * Write a buffer to a temporary file. Returns { tmpPath, cleanup }.
+ * Caller MUST call cleanup() when done.
+ */
+function bufferToTempFile(buffer, ext = '.bin') {
+  const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'refboard-pdf-'));
+  const tmpPath = path.join(tmpDir, `file${ext}`);
+  fs.writeFileSync(tmpPath, buffer);
+  return {
+    tmpPath,
+    cleanup() {
+      try { fs.unlinkSync(tmpPath); } catch {}
+      try { fs.rmdirSync(tmpDir); } catch {}
+    },
+  };
+}
+
+/**
+ * Extract PDF metadata via pdfinfo.
+ * Returns { pageCount, dimensions: [{ w, h }, ...] }
+ */
+async function pdfInfo(filePath) {
+  // Get basic info (page count)
+  const { stdout: basicOut } = await execFileAsync('pdfinfo', [filePath], { timeout: TIMEOUT_MS });
+
+  let pageCount = 0;
+  for (const line of basicOut.split('\n')) {
+    const match = line.match(/^Pages:\s+(\d+)/);
+    if (match) {
+      pageCount = parseInt(match[1], 10);
+      break;
+    }
+  }
+
+  if (pageCount === 0) {
+    throw new Error('Could not determine PDF page count');
+  }
+
+  // Get per-page dimensions
+  const dimensions = [];
+  const { stdout: dimOut } = await execFileAsync(
+    'pdfinfo',
+    ['-f', '1', '-l', String(pageCount), filePath],
+    { timeout: TIMEOUT_MS }
+  );
+
+  // Parse "Page N size: W x H pts" lines
+  for (const line of dimOut.split('\n')) {
+    const match = line.match(/^Page\s+\d+\s+size:\s+([\d.]+)\s+x\s+([\d.]+)/);
+    if (match) {
+      dimensions.push({
+        w: Math.round(parseFloat(match[1])),
+        h: Math.round(parseFloat(match[2])),
+      });
+    }
+  }
+
+  // Fallback: if no per-page sizes found, try global "Page size:"
+  if (dimensions.length === 0) {
+    for (const line of basicOut.split('\n')) {
+      const match = line.match(/^Page size:\s+([\d.]+)\s+x\s+([\d.]+)/);
+      if (match) {
+        const dim = {
+          w: Math.round(parseFloat(match[1])),
+          h: Math.round(parseFloat(match[2])),
+        };
+        for (let i = 0; i < pageCount; i++) {
+          dimensions.push(dim);
+        }
+        break;
+      }
+    }
+  }
+
+  return { pageCount, dimensions };
+}
+
+/**
+ * Render a single PDF page to PNG at the given DPI.
+ * Returns a PNG Buffer.
+ */
+async function pdfRenderPage(filePath, pageNum, dpi = 150) {
+  const outDir = fs.mkdtempSync(path.join(os.tmpdir(), 'refboard-render-'));
+
+  try {
+    const outPrefix = path.join(outDir, 'page');
+
+    await execFileAsync('pdftoppm', [
+      '-png',
+      '-r', String(dpi),
+      '-f', String(pageNum),
+      '-l', String(pageNum),
+      filePath,
+      outPrefix,
+    ], { timeout: TIMEOUT_MS });
+
+    // pdftoppm names output like page-01.png, page-1.png, etc. — discover it
+    const files = fs.readdirSync(outDir).filter(f => f.endsWith('.png'));
+    if (files.length === 0) {
+      throw new Error(`pdftoppm produced no output for page ${pageNum}`);
+    }
+
+    const pngPath = path.join(outDir, files[0]);
+    const pngBuffer = fs.readFileSync(pngPath);
+    return pngBuffer;
+  } finally {
+    // Clean up temp dir
+    try {
+      for (const f of fs.readdirSync(outDir)) {
+        fs.unlinkSync(path.join(outDir, f));
+      }
+      fs.rmdirSync(outDir);
+    } catch {}
+  }
+}
+
+module.exports = { pdfInfo, pdfRenderPage, bufferToTempFile };