diff --git a/tests/test_web.py b/tests/test_web.py
index 364edd5..5a4c3c5 100644
--- a/tests/test_web.py
+++ b/tests/test_web.py
@@ -1,4 +1,5 @@
 from datetime import datetime
+
 import pytest
 from starlette.testclient import TestClient
 
diff --git a/unwind/db.py b/unwind/db.py
index b91d3c1..34abd81 100644
--- a/unwind/db.py
+++ b/unwind/db.py
@@ -18,7 +18,9 @@ from .models import (
     asplain,
     fields,
     fromplain,
+    movies,
     optional_fields,
+    ratings,
     utcnow,
 )
 from .types import ULID
@@ -427,77 +429,72 @@ async def find_ratings(
     limit_rows: int = 10,
     user_ids: Iterable[str] = [],
 ):
-    values: dict[str, int | str] = {
-        "limit_rows": limit_rows,
-    }
-
     conditions = []
+
     if title:
-        values["escape"] = "#"
-        escaped_title = sql_escape(title, char=values["escape"])
-        values["pattern"] = (
+        escape_char = "#"
+        escaped_title = sql_escape(title, char=escape_char)
+        pattern = (
             "_".join(escaped_title.split())
             if exact
             else "%" + "%".join(escaped_title.split()) + "%"
         )
         conditions.append(
-            f"""
-            (
-                {Movie._table}.title LIKE :pattern ESCAPE :escape
-                OR {Movie._table}.original_title LIKE :pattern ESCAPE :escape
+            sa.or_(
+                movies.c.title.like(pattern, escape=escape_char),
+                movies.c.original_title.like(pattern, escape=escape_char),
             )
-            """
         )
 
-    if yearcomp:
-        op, year = yearcomp
-        assert op in "<=>"
-        values["year"] = year
-        conditions.append(f"{Movie._table}.release_year{op}:year")
+    match yearcomp:
+        case ("<", year):
+            conditions.append(movies.c.release_year < year)
+        case ("=", year):
+            conditions.append(movies.c.release_year == year)
+        case (">", year):
+            conditions.append(movies.c.release_year > year)
 
-    if media_type:
-        values["media_type"] = media_type
-        conditions.append(f"{Movie._table}.media_type=:media_type")
+    if media_type is not None:
+        conditions.append(movies.c.media_type == media_type)
 
     if ignore_tv_episodes:
-        conditions.append(f"{Movie._table}.media_type!='TV Episode'")
+        conditions.append(movies.c.media_type != "TV Episode")
 
-    user_condition = "1=1"
+    user_condition = []
     if user_ids:
-        uvs = {f"user_id_{i}": v for i, v in enumerate(user_ids, start=1)}
-        values.update(uvs)
-        user_condition = f"{Rating._table}.user_id IN ({','.join(':'+n for n in uvs)})"
+        user_condition.append(ratings.c.user_id.in_(user_ids))
 
-    query = f"""
-        SELECT DISTINCT {Rating._table}.movie_id
-        FROM {Rating._table}
-        LEFT JOIN {Movie._table} ON {Movie._table}.id={Rating._table}.movie_id
-        WHERE {user_condition}{(' AND ' + ' AND '.join(conditions)) if conditions else ''}
-        ORDER BY length({Movie._table}.title) ASC, {Rating._table}.rating_date DESC, {Movie._table}.imdb_score DESC
-        LIMIT :limit_rows
-    """
+    query = (
+        sa.select(ratings.c.movie_id)
+        .distinct()
+        .outerjoin_from(ratings, movies, movies.c.id == ratings.c.movie_id)
+        .where(*conditions, *user_condition)
+        .order_by(
+            sa.func.length(movies.c.title).asc(),
+            ratings.c.rating_date.desc(),
+            movies.c.imdb_score.desc(),
+        )
+        .limit(limit_rows)
+    )
     async with locked_connection() as conn:
-        rows = await conn.fetch_all(bindparams(query, values))
-        movie_ids = tuple(r._mapping["movie_id"] for r in rows)
+        rows = conn.iterate(query)
+        movie_ids = [r.movie_id async for r in rows]
 
     if include_unrated and len(movie_ids) < limit_rows:
-        sqlin, sqlin_vals = sql_in("id", movie_ids, not_=True)
-        query = f"""
-            SELECT DISTINCT id AS movie_id
-            FROM {Movie._table}
-            WHERE {sqlin}
-            {('AND ' + ' AND '.join(conditions)) if conditions else ''}
-            ORDER BY length(title) ASC, imdb_score DESC, release_year DESC
-            LIMIT :limit_rows
-        """
-        async with locked_connection() as conn:
-            rows = await conn.fetch_all(
-                bindparams(
-                    query,
-                    {**values, **sqlin_vals, "limit_rows": limit_rows - len(movie_ids)},
-                )
+        query = (
+            sa.select(movies.c.id.label("movie_id"))
+            .distinct()
+            .where(movies.c.id.not_in(movie_ids), *conditions)
+            .order_by(
+                sa.func.length(movies.c.title).asc(),
+                movies.c.imdb_score.desc(),
+                movies.c.release_year.desc(),
             )
-            movie_ids += tuple(r._mapping["movie_id"] for r in rows)
+            .limit(limit_rows - len(movie_ids))
+        )
+        async with locked_connection() as conn:
+            rows = conn.iterate(query)
+            movie_ids += [r.movie_id async for r in rows]
 
     return await ratings_for_movie_ids(ids=movie_ids)
 
@@ -507,9 +504,6 @@ async def ratings_for_movie_ids(
 ) -> Iterable[dict[str, Any]]:
     conds = []
 
-    ratings = Rating.__table__
-    movies = Movie.__table__
-
     if ids:
         conds.append(movies.c.id.in_([str(x) for x in ids]))
 
diff --git a/unwind/models.py b/unwind/models.py
index 964ac01..6d40b35 100644
--- a/unwind/models.py
+++ b/unwind/models.py
@@ -12,8 +12,8 @@ from typing import (
     Literal,
     Mapping,
     Type,
-    TypeVar,
     TypedDict,
+    TypeVar,
     Union,
     get_args,
     get_origin,
@@ -318,6 +318,8 @@ class Movie:
         self._is_lazy = False
 
 
+movies = Movie.__table__
+
 _RelationSentinel = object()
 """Mark a model field as containing external data.
 
@@ -372,6 +374,8 @@ class Rating:
         )
 
 
+ratings = Rating.__table__
+
 Access = Literal[
     "r",  # read
     "i",  # index