diff --git a/pipeline_runner/processor.py b/pipeline_runner/processor.py
new file mode 100644
index 0000000..6f31bf9
--- /dev/null
+++ b/pipeline_runner/processor.py
@@ -0,0 +1,174 @@
+import csv
+import gzip
+import json
+import math
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+
+from .config import LTTB_MAX_PTS
+
+
+def _lttb(points: list[dict], max_pts: int) -> list[dict]:
+    """Largest Triangle Three Buckets downsampling."""
+    n = len(points)
+    if n <= max_pts:
+        return points
+    ts = np.array([p["t"] for p in points], dtype=float)
+    vs = np.array([p["v"] if isinstance(p["v"], (int, float)) else 0.0 for p in points], dtype=float)
+    bucket_size = (n - 2) / (max_pts - 2)
+    out = [points[0]]
+    a = 0
+    for i in range(max_pts - 2):
+        avg_start = int((i + 1) * bucket_size) + 1
+        avg_end = min(int((i + 2) * bucket_size) + 1, n)
+        avg_t = np.mean(ts[avg_start:avg_end])
+        avg_v = np.mean(vs[avg_start:avg_end])
+        rng_start = int(i * bucket_size) + 1
+        rng_end = min(int((i + 1) * bucket_size) + 1, n)
+        max_area = -1.0
+        best = rng_start
+        at, av = ts[a], vs[a]
+        for j in range(rng_start, rng_end):
+            area = abs((at - avg_t) * (vs[j] - av) - (ts[j] - at) * (avg_v - av)) * 0.5
+            if area > max_area:
+                max_area = area
+                best = j
+        out.append(points[best])
+        a = best
+    out.append(points[-1])
+    return out
+
+
+def _ts_to_epoch(ts_str: str) -> float:
+    """Parse 'YYYY-MM-DD HH:MM:SS.ffffff' → epoch seconds."""
+    for fmt in ("%Y-%m-%d %H:%M:%S.%f", "%Y-%m-%d %H:%M:%S"):
+        try:
+            dt = datetime.strptime(ts_str.strip(), fmt).replace(tzinfo=timezone.utc)
+            return dt.timestamp()
+        except ValueError:
+            continue
+    return 0.0
+
+
+def _read_nav_log(nav_log_path: Path) -> dict[str, list[dict]]:
+    """Read long-format navigation_log.csv → dict of signal arrays."""
+    signals: dict[str, list] = {}
+    wanted = {"Yaw", "Heading", "Roll", "Pitch", "BattVoltage", "gps_fix",
+              "Armed", "Mode", "M1", "M2"}
+    with open(nav_log_path, newline="", encoding="utf-8") as f:
+        reader = csv.reader(f)
+        next(reader, None)  # skip header
+        for row in reader:
+            if len(row) < 3:
+                continue
+            ts_str, field, val = row[0], row[1], row[2]
+            if field not in wanted:
+                continue
+            t = _ts_to_epoch(ts_str)
+            try:
+                v: Any = float(val)
+            except ValueError:
+                v = val.strip()
+            signals.setdefault(field, []).append({"t": t, "v": v})
+    return signals
+
+
+def _read_usbl_csv(usbl_csv_path: Path) -> dict[str, list[dict]]:
+    """Read combined_usbl.csv → Dist and Azimuth signal arrays."""
+    dist_pts, az_pts = [], []
+    with open(usbl_csv_path, newline="", encoding="utf-8") as f:
+        reader = csv.DictReader(f)
+        for row in reader:
+            try:
+                t = _ts_to_epoch(row["Timestamp"])
+                d = float(row["Dist"]) if row["Dist"] else None
+                az = float(row["Azimuth"]) if row["Azimuth"] else None
+            except (KeyError, ValueError):
+                continue
+            if d is not None and not math.isnan(d):
+                dist_pts.append({"t": t, "v": d})
+            if az is not None and not math.isnan(az):
+                az_pts.append({"t": t, "v": az})
+    return {"usbl_dist": dist_pts, "usbl_angle": az_pts}
+
+
+def _read_mcap_signals(mcap_json_path: Path) -> dict[str, list[dict]]:
+    """Read mcap_signals.json → depth, motors M1-M6, state signals."""
+    with open(mcap_json_path) as f:
+        data = json.load(f)
+    signals: dict[str, list[dict]] = {}
+
+    def _unpack(key: str, series: list):
+        pts = []
+        for item in series:
+            t = item.get("t_ms", item.get("t", 0))
+            if isinstance(t, (int, float)) and t > 1e9:
+                t = t / 1000.0  # ms → s
+            pts.append({"t": float(t), "v": item.get("v", item.get("value", 0))})
+        signals[key] = pts
+
+    if "depth" in data:
+        _unpack("depth", data["depth"])
+    if "pwm_auv" in data:
+        for m_key, series in data["pwm_auv"].items():
+            _unpack(m_key.lower(), series)
+    if "state" in data:
+        _unpack("arm_status", data["state"])
+    # New signals added by extended extract_mcap_signals.py
+    for key in ("pitch", "roll", "yaw", "altitude", "battery_v", "obstacle_dist"):
+        if key in data:
+            _unpack(key, data[key])
+    return signals
+
+
+def write_usv_json(
+    nav_log_path: Path,
+    usbl_csv_path: Path | None,
+    output_path: Path,
+    sortie_id: str,
+    date: str,
+) -> None:
+    signals = _read_nav_log(nav_log_path)
+    if usbl_csv_path and usbl_csv_path.exists():
+        signals.update(_read_usbl_csv(usbl_csv_path))
+
+    t_all = [p["t"] for pts in signals.values() for p in pts if pts]
+    meta = {
+        "sortie": sortie_id,
+        "date": date,
+        "vehicle": "USV",
+        "t_start": min(t_all) if t_all else 0,
+        "t_end": max(t_all) if t_all else 0,
+    }
+
+    downsampled = {k: _lttb(v, LTTB_MAX_PTS) for k, v in signals.items()}
+    payload = json.dumps({"meta": meta, "signals": downsampled}).encode()
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    with gzip.open(output_path, "wb") as f:
+        f.write(payload)
+
+
+def write_auv_json(
+    mcap_json_path: Path,
+    output_path: Path,
+    auv_id: str,
+    sortie_id: str,
+    date: str,
+) -> None:
+    signals = _read_mcap_signals(mcap_json_path)
+    t_all = [p["t"] for pts in signals.values() for p in pts if pts]
+    meta = {
+        "sortie": sortie_id,
+        "date": date,
+        "vehicle": auv_id,
+        "t_start": min(t_all) if t_all else 0,
+        "t_end": max(t_all) if t_all else 0,
+    }
+    downsampled = {k: _lttb(v, LTTB_MAX_PTS) for k, v in signals.items()}
+    payload = json.dumps({"meta": meta, "signals": downsampled}).encode()
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    with gzip.open(output_path, "wb") as f:
+        f.write(payload)