diff --git a/scripts/build-epub.py b/scripts/build-epub.py
new file mode 100644
index 00000000..f7edaa25
--- /dev/null
+++ b/scripts/build-epub.py
@@ -0,0 +1,299 @@
+#!/usr/bin/env python3
+"""Build an EPUB from the Algorithmica HPC book content.
+
+Usage:
+    python3 build-epub.py [--lang en|ru] [--include-drafts] [--output FILE]
+
+Requirements:
+    - Python 3.8+
+    - pandoc (https://pandoc.org/installing.html)
+
+The script reads Hugo content directories, orders pages by their front matter
+`weight` field, strips YAML front matter, resolves image paths, and invokes
+pandoc to produce an EPUB3 file with MathML for LaTeX rendering.
+"""
+
+import argparse
+import os
+import re
+import subprocess
+import sys
+import tempfile
+import shutil
+from pathlib import Path
+
+BOOK_CONFIGS = {
+    "en": {
+        "content_dir": "content/english/hpc",
+        "title": "Algorithms for Modern Hardware",
+        "author": "Sergey Slotin",
+        "lang": "en-US",
+        "output": "algorithms-for-modern-hardware.epub",
+    },
+    "ru": {
+        "content_dir": "content/russian/cs",
+        "title": "Алгоритмика",
+        "author": "Sergey Slotin",
+        "lang": "ru-RU",
+        "output": "algoritmika.epub",
+    },
+}
+
+FRONT_MATTER_RE = re.compile(r"^---\s*\n.*?\n---\s*\n", re.DOTALL)
+SKIP_CHAPTERS = {"slides"}
+
+
+def parse_front_matter(text):
+    """Extract YAML front matter fields without a YAML dependency."""
+    match = FRONT_MATTER_RE.match(text)
+    if not match:
+        return {}, text
+    fm_block = match.group(0)
+    body = text[match.end():]
+    meta = {}
+    for line in fm_block.split("\n"):
+        if ":" in line and not line.startswith("---"):
+            key, _, value = line.partition(":")
+            meta[key.strip()] = value.strip().strip('"').strip("'")
+    return meta, body
+
+
+def get_weight(meta):
+    try:
+        return int(meta.get("weight", "0"))
+    except ValueError:
+        return 0
+
+
+def is_draft(meta):
+    return meta.get("draft", "").lower() == "true"
+
+
+def collect_chapters(content_dir):
+    """Walk the content directory and return ordered list of (chapter_meta, chapter_dir, pages)."""
+    content_path = Path(content_dir)
+    chapters = []
+
+    for item in content_path.iterdir():
+        if not item.is_dir():
+            if item.suffix == ".md" and item.name != "_index.md":
+                text = item.read_text(encoding="utf-8")
+                meta, body = parse_front_matter(text)
+                chapters.append((meta, None, [(meta, item)]))
+            continue
+
+        if item.name in SKIP_CHAPTERS or item.name == "img":
+            continue
+
+        index_file = item / "_index.md"
+        if not index_file.exists():
+            continue
+
+        index_text = index_file.read_text(encoding="utf-8")
+        chapter_meta, chapter_body = parse_front_matter(index_text)
+
+        pages = []
+        for md_file in sorted(item.glob("*.md")):
+            if md_file.name == "_index.md":
+                if chapter_body.strip():
+                    pages.append((chapter_meta, md_file))
+                continue
+            page_text = md_file.read_text(encoding="utf-8")
+            page_meta, _ = parse_front_matter(page_text)
+            pages.append((page_meta, md_file))
+
+        pages.sort(key=lambda p: get_weight(p[0]))
+        chapters.append((chapter_meta, item, pages))
+
+    chapters.sort(key=lambda c: get_weight(c[0]))
+    return chapters
+
+
+def resolve_image_path(match, md_file, repo_root):
+    """Resolve a relative image path to an absolute filesystem path.
+
+    Hugo serves pages at /chapter/page/ (extra directory level), so ../img/foo.png
+    from a page actually means the sibling img/ directory in the same chapter folder.
+    On the filesystem, ../img/ from chapter/page.md goes one level too high, so we
+    handle this Hugo-specific path resolution.
+    """
+    alt = match.group(1)
+    img_path = match.group(2)
+
+    if img_path.startswith("http://") or img_path.startswith("https://"):
+        return match.group(0)
+
+    md_dir = md_file.parent
+
+    # Direct filesystem resolution first
+    resolved = (md_dir / img_path).resolve()
+    if resolved.exists():
+        return f"![{alt}]({resolved})"
+
+    # Hugo path fix: ../img/X from chapter/page.md means chapter/img/X
+    # because Hugo renders as /chapter/page/ (adds a virtual directory level)
+    if img_path.startswith("../img/"):
+        filename = img_path[len("../img/"):]
+        hugo_resolved = (md_dir / "img" / filename).resolve()
+        if hugo_resolved.exists():
+            return f"![{alt}]({hugo_resolved})"
+
+    # Try static/img/ as a fallback (Hugo's static asset directory)
+    static_path = (Path(repo_root) / "static" / img_path.lstrip("/")).resolve()
+    if static_path.exists():
+        return f"![{alt}]({static_path})"
+
+    # Try static/img/<filename> as last resort
+    filename = Path(img_path).name
+    static_fallback = (Path(repo_root) / "static" / "img" / filename).resolve()
+    if static_fallback.exists():
+        return f"![{alt}]({static_fallback})"
+
+    # Return absolute path even if missing (pandoc will warn but continue)
+    return f"![{alt}]({resolved})"
+
+
+def process_page(md_file, repo_root, include_drafts):
+    """Read a page, strip front matter, resolve images, return processed markdown."""
+    text = md_file.read_text(encoding="utf-8")
+    meta, body = parse_front_matter(text)
+
+    if not include_drafts and is_draft(meta):
+        return None, meta
+
+    img_re = re.compile(r"!\[([^\]]*)\]\(([^)]+)\)")
+    body = img_re.sub(lambda m: resolve_image_path(m, md_file, repo_root), body)
+
+    # Remove HTML comments
+    body = re.sub(r"<!--.*?-->", "", body, flags=re.DOTALL)
+
+    return body, meta
+
+
+def build_epub(config, include_drafts, output_path, repo_root):
+    content_dir = Path(config["content_dir"])
+    if not content_dir.exists():
+        print(f"Error: content directory '{content_dir}' not found.", file=sys.stderr)
+        print("Run this script from the repository root.", file=sys.stderr)
+        sys.exit(1)
+
+    chapters = collect_chapters(content_dir)
+
+    with tempfile.TemporaryDirectory() as tmpdir:
+        combined_path = Path(tmpdir) / "book.md"
+        page_count = 0
+
+        with open(combined_path, "w", encoding="utf-8") as out:
+            for chapter_meta, chapter_dir, pages in chapters:
+                if not include_drafts and is_draft(chapter_meta):
+                    continue
+
+                chapter_title = chapter_meta.get("title", "")
+
+                first_page = True
+                for page_meta, md_file in pages:
+                    body, meta = process_page(md_file, repo_root, include_drafts)
+                    if body is None:
+                        continue
+
+                    page_title = meta.get("title", "")
+
+                    if first_page and md_file.name == "_index.md":
+                        out.write(f"\n# {chapter_title}\n\n")
+                        if body.strip():
+                            out.write(body)
+                            out.write("\n\n")
+                        first_page = False
+                    elif first_page and chapter_title:
+                        out.write(f"\n# {chapter_title}\n\n")
+                        if page_title and page_title != chapter_title:
+                            out.write(f"## {page_title}\n\n")
+                        out.write(body)
+                        out.write("\n\n")
+                        first_page = False
+                    else:
+                        if page_title:
+                            out.write(f"## {page_title}\n\n")
+                        out.write(body)
+                        out.write("\n\n")
+
+                    page_count += 1
+
+        if page_count == 0:
+            print("Error: no pages found to include.", file=sys.stderr)
+            sys.exit(1)
+
+        print(f"Collected {page_count} pages across {len(chapters)} sections.")
+
+        cmd = [
+            "pandoc",
+            str(combined_path),
+            "--from", "markdown+tex_math_dollars+pipe_tables+fenced_code_blocks",
+            "--to", "epub3",
+            "--mathml",
+            "--toc",
+            "--toc-depth=2",
+            "--metadata", f"title={config['title']}",
+            "--metadata", f"author={config['author']}",
+            "--metadata", f"lang={config['lang']}",
+            "--epub-chapter-level=1",
+            "--wrap=none",
+            "-o", str(output_path),
+        ]
+
+        cover = Path(repo_root) / "static" / "img" / "cover.png"
+        if not cover.exists():
+            cover = Path(repo_root) / "static" / "img" / "logo.png"
+        if cover.exists():
+            cmd.extend(["--epub-cover-image", str(cover)])
+
+        print("Running pandoc...")
+        result = subprocess.run(cmd, capture_output=True, text=True)
+
+        if result.returncode != 0:
+            print("pandoc stderr:", result.stderr, file=sys.stderr)
+            sys.exit(result.returncode)
+
+        if result.stderr:
+            warnings = [l for l in result.stderr.strip().split("\n") if l.strip()]
+            if warnings:
+                print(f"  ({len(warnings)} warnings from pandoc)")
+
+        file_size = os.path.getsize(output_path)
+        print(f"Done: {output_path} ({file_size / 1024 / 1024:.1f} MB)")
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Build an EPUB from Algorithmica book content."
+    )
+    parser.add_argument(
+        "--lang",
+        choices=["en", "ru"],
+        default="en",
+        help="Which edition to build (default: en)",
+    )
+    parser.add_argument(
+        "--include-drafts",
+        action="store_true",
+        help="Include pages marked as draft",
+    )
+    parser.add_argument(
+        "--output", "-o",
+        help="Output file path (default: based on language)",
+    )
+    args = parser.parse_args()
+
+    config = BOOK_CONFIGS[args.lang]
+    output_path = Path(args.output) if args.output else Path(config["output"])
+
+    if not shutil.which("pandoc"):
+        print("Error: pandoc not found. Install it: https://pandoc.org/installing.html", file=sys.stderr)
+        sys.exit(1)
+
+    repo_root = Path(__file__).parent.resolve()
+    build_epub(config, args.include_drafts, output_path, repo_root)
+
+
+if __name__ == "__main__":
+    main()