diff --git a/.DS_Store b/.DS_Store
new file mode 100644
index 0000000..3618cab
Binary files /dev/null and b/.DS_Store differ
diff --git a/.gitignore b/.gitignore
index f85f30e..47969bd 100644
--- a/.gitignore
+++ b/.gitignore
@@ -215,3 +215,5 @@ __marimo__/
 
 # Streamlit
 .streamlit/secrets.toml
+
+mimic-iv/
diff --git a/group_code/__init__.py b/group_code/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/group_code/helper.py b/group_code/helper.py
new file mode 100644
index 0000000..f2e5854
--- /dev/null
+++ b/group_code/helper.py
@@ -0,0 +1,21 @@
+# gpt generated code
+def combine_dataframes(dfs, tags, on="subject_id"):
+    """
+    Combine multiple dataframes on a common column, prefixing each dataframe's
+    columns (except the merge column) with a given tag.
+    Keeps only matched IDs across all dataframes (inner join).
+    """
+    assert len(dfs) == len(tags), "Number of dataframes must match number of tags"
+
+    # Rename columns in each dataframe (except merge key)
+    renamed_dfs = []
+    for df, tag in zip(dfs, tags):
+        df_renamed = df.rename(columns={col: f"{tag}_{col}" for col in df.columns if col != on})
+        renamed_dfs.append(df_renamed)
+
+    # Iteratively merge with inner join
+    combined = renamed_dfs[0]
+    for df in renamed_dfs[1:]:
+        combined = combined.merge(df, on=on, how="inner")
+
+    return combined
diff --git a/group_code/mm_dataset.py b/group_code/mm_dataset.py
new file mode 100644
index 0000000..5bf72e3
--- /dev/null
+++ b/group_code/mm_dataset.py
@@ -0,0 +1,46 @@
+from group_code.helper import combine_dataframes
+from group_code.uni_model.ecg import ECG_uni
+from group_code.uni_model.echo import ECHO_uni
+from group_code.uni_model.ehr import EHR_uni
+from mmai25_hackathon.dataset import BaseDataset
+
+
+class EEE_dataset(BaseDataset):
+    def __init__(self, data_roots, data_mods):
+        self.uni_dict = {}
+        for mod, r_path in zip(data_mods, data_roots):
+            if mod == "ecg":
+                self.uni_dict[mod] = ECG_uni(r_path)
+            elif mod == "ehr":
+                self.uni_dict[mod] = EHR_uni(r_path)
+            elif mod == "echo":
+                self.uni_dict[mod] = ECHO_uni(r_path)
+            else:
+                print("Modality not supported.")
+
+        self.combined_records = combine_dataframes(
+            [val.records for key, val in self.uni_dict.items() if key != "ehr"],
+            [key for key, val in self.uni_dict.items() if key != "ehr"],
+        )
+
+    def __len__(self) -> int:
+        return len(self.combined_records)
+
+    def __getitem__(self, idx: int):
+        subject_id = self.combined_records.iloc[idx]["subject_id"]
+        return_dict = {}
+        for key, val in self.uni_dict.items():
+            return_dict[key] = val.fetch(subject_id)
+        return return_dict
+
+    def __add__(self, data_roots, data_mods):
+        for mod, r_path in zip(data_mods, data_roots):
+            if mod == "ecg":
+                self.uni_dict[mod] = ECG_uni(r_path)
+        self.combined_records = combine_dataframes(
+            [val.records for key, val in self.uni_dict.items() if key != "ehr"],
+            [key for key, val in self.uni_dict.items() if key != "ehr"],
+        )
+
+    def get_idx_from_sub_id(self, subject_id):
+        return self.combined_records.index[self.combined_records["subject_id"] == subject_id][0]
diff --git a/group_code/uni_model/ecg.py b/group_code/uni_model/ecg.py
new file mode 100644
index 0000000..5e3411b
--- /dev/null
+++ b/group_code/uni_model/ecg.py
@@ -0,0 +1,24 @@
+from torch.utils.data import Dataset
+
+from mmai25_hackathon.load_data.ecg import load_ecg_record, load_mimic_iv_ecg_record_list
+
+
+class ECG_uni(Dataset):
+    def __init__(self, mod_root):
+        self.records = load_mimic_iv_ecg_record_list(mod_root)
+
+    def __len__(self) -> int:
+        return len(self.records)
+
+    def __getitem__(self, idx: int):
+        subject_id = self.records.iloc[idx]["subject_id"]
+        sig, fields = load_ecg_record(self.records.iloc[idx]["hea_path"])
+
+        return {subject_id: [sig, fields]}
+
+    def get_idx_by_subject(self, subject_id):
+        return self.records.index[self.records["subject_id"] == subject_id][0]
+
+    def fetch(self, subject_id):
+        idx = self.get_idx_by_subject(subject_id)
+        return self.__getitem__(idx)
diff --git a/group_code/uni_model/echo.py b/group_code/uni_model/echo.py
new file mode 100644
index 0000000..bc58e6b
--- /dev/null
+++ b/group_code/uni_model/echo.py
@@ -0,0 +1,27 @@
+# noqa: F403
+from torch.utils.data import Dataset
+
+from mmai25_hackathon.load_data.echo import load_echo_dicom, load_mimic_iv_echo_record_list
+
+
+class ECHO_uni(Dataset):
+    def __init__(self, mod_root):
+        self.records = load_mimic_iv_echo_record_list(mod_root)
+
+    def __len__(self) -> int:
+        return len(self.records)
+
+    def __getitem__(self, idx: int):
+        subject_id = self.records.iloc[idx]["subject_id"]
+        frames, meta = load_echo_dicom(self.records.iloc[idx]["echo_path"])
+        meta_filtered = {
+            k: meta[k] for k in ("NumberOfFrames", "Rows", "Columns", "FrameTime", "CineRate") if k in meta
+        }
+        return {subject_id: meta_filtered}
+
+    def get_idx_by_subject(self, subject_id):
+        return self.records.index[self.records["subject_id"] == subject_id][0]
+
+    def fetch(self, subject_id):
+        idx = self.get_idx_by_subject(subject_id)
+        return self.__getitem__(idx)
diff --git a/group_code/uni_model/ehr.py b/group_code/uni_model/ehr.py
new file mode 100644
index 0000000..69c3256
--- /dev/null
+++ b/group_code/uni_model/ehr.py
@@ -0,0 +1,30 @@
+from torch.utils.data import Dataset
+
+from mmai25_hackathon.load_data.ehr import load_mimic_iv_ehr
+
+
+class EHR_uni(Dataset):
+    def __init__(self, mod_root):
+        self.root = mod_root
+
+    def __len__(self) -> int:
+        return len(self.records)
+
+    def __getitem__(self, subject_id):
+        return self.fetch(subject_id)
+
+    def fetch(self, subject_id):
+        dfs_new = load_mimic_iv_ehr(
+            ehr_path=self.root,
+            module="both",
+            tables=["icustays", "admissions"],
+            index_cols=["subject_id", "hadm_id"],
+            subset_cols={
+                "icustays": ["first_careunit"],
+                "admissions": ["admittime"],
+            },
+            filter_rows={"subject_id": [int(subject_id)]},
+            merge=True,
+            join="inner",
+        )
+        return {subject_id: dfs_new}
diff --git a/group_test/test_102.py b/group_test/test_102.py
new file mode 100644
index 0000000..5af8cfb
--- /dev/null
+++ b/group_test/test_102.py
@@ -0,0 +1,32 @@
+from group_code.mm_dataset import EEE_dataset
+from group_code.uni_model.ecg import ECG_uni
+from group_code.uni_model.echo import ECHO_uni
+from group_code.uni_model.ehr import EHR_uni
+
+ecg_root = "mimic-iv/mimic-iv-ecg-diagnostic-electrocardiogram-matched-subset-1.0/"
+echo_root = "mimic-iv/mimic-iv-echo-0.1.physionet.org/"
+ehr_root = "mimic-iv/mimic-iv-3.1/"
+
+
+def test_by_id(id):
+    ds = EEE_dataset([ecg_root, ehr_root, echo_root], ["ecg", "ehr", "echo"])
+
+    ds_idx = ds.get_idx_from_sub_id(id)
+    full_results = ds[ds_idx]
+
+    ecg = ECG_uni(ecg_root)
+    ehr = EHR_uni(ehr_root)
+    echo = ECHO_uni(echo_root)
+
+    ecg_res = ecg.fetch(id)
+    ehr_res = ehr.fetch(id)
+    echo_res = echo.fetch(id)
+
+    assert full_results["ecg"][id][1]["comments"] == ecg_res[id][1]["comments"]
+    assert full_results["echo"][id]["Rows"] == echo_res[id]["Rows"]
+    assert len(full_results["ehr"][id]) == len(ehr_res[id])
+
+
+if __name__ == "__main__":
+    test_by_id(102)
+    test_by_id(101)
diff --git a/try.ipynb b/try.ipynb
new file mode 100644
index 0000000..6684705
--- /dev/null
+++ b/try.ipynb
@@ -0,0 +1,248 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "f6d69c1a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "ecg_root = 'mimic-iv/mimic-iv-ecg-diagnostic-electrocardiogram-matched-subset-1.0/'\n",
+    "echo_root = 'mimic-iv/mimic-iv-echo-0.1.physionet.org/'\n",
+    "ehr_root = 'mimic-iv/mimic-iv-3.1/'"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "c391dde8",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from mmai25_hackathon.load_data.ecg import * \n",
+    "from mmai25_hackathon.load_data.echo import * \n",
+    "from mmai25_hackathon.load_data.ehr import * "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "e6cf19b4",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "record = load_mimic_iv_ecg_record_list(ecg_root)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "2649c7ad",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from group_code.mm_dataset import *"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "221279ec",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "{'ecg': {np.int64(101): [array([[ 0.02 ,  0.06 ,  0.04 , ...,  0.035,  0.42 , -0.28 ],\n",
+       "          [ 0.   ,  0.075,  0.075, ...,  0.045,  0.41 , -0.295],\n",
+       "          [-0.015,  0.075,  0.09 , ...,  0.055,  0.4  , -0.3  ],\n",
+       "          ...,\n",
+       "          [-0.015, -0.015,  0.   , ..., -0.235,  0.34 , -0.065],\n",
+       "          [ 0.   , -0.015, -0.015, ..., -0.235,  0.34 , -0.065],\n",
+       "          [ 0.   , -0.02 , -0.02 , ..., -0.225,  0.34 , -0.065]],\n",
+       "         shape=(5000, 12)),\n",
+       "   {'fs': 500,\n",
+       "    'sig_len': 5000,\n",
+       "    'n_sig': 12,\n",
+       "    'base_date': datetime.date(2180, 8, 6),\n",
+       "    'base_time': datetime.time(9, 7),\n",
+       "    'units': ['mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV'],\n",
+       "    'sig_name': ['I',\n",
+       "     'II',\n",
+       "     'III',\n",
+       "     'aVR',\n",
+       "     'aVF',\n",
+       "     'aVL',\n",
+       "     'V1',\n",
+       "     'V2',\n",
+       "     'V3',\n",
+       "     'V4',\n",
+       "     'V5',\n",
+       "     'V6'],\n",
+       "    'comments': ['<subject_id>: 101']}]},\n",
+       " 'ehr': {np.int64(101):    subject_id  hadm_id         admittime  \\\n",
+       "  0         101        1  24/02/2196 14:38   \n",
+       "  1         101        2  17/09/2153 17:08   \n",
+       "  2         101        3  18/08/2134 02:02   \n",
+       "  \n",
+       "                                      first_careunit  \n",
+       "  0                                   Neuro Stepdown  \n",
+       "  1  Neuro Surgical Intensive Care Unit (Neuro SICU)  \n",
+       "  2                               Neuro Intermediate  },\n",
+       " 'echo': {np.int64(101): {'NumberOfFrames': '58',\n",
+       "   'Rows': 708,\n",
+       "   'Columns': 1016,\n",
+       "   'FrameTime': '33.6842',\n",
+       "   'CineRate': '30'}}}"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "ds = EEE_dataset(\n",
+    "    [ecg_root, ehr_root, echo_root],\n",
+    "    ['ecg', 'ehr', 'echo']\n",
+    ")\n",
+    "ds[4]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "7b9466a3",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "{'ecg': {np.int64(101): [array([[ 0.02 ,  0.06 ,  0.04 , ...,  0.035,  0.42 , -0.28 ],\n",
+       "          [ 0.   ,  0.075,  0.075, ...,  0.045,  0.41 , -0.295],\n",
+       "          [-0.015,  0.075,  0.09 , ...,  0.055,  0.4  , -0.3  ],\n",
+       "          ...,\n",
+       "          [-0.015, -0.015,  0.   , ..., -0.235,  0.34 , -0.065],\n",
+       "          [ 0.   , -0.015, -0.015, ..., -0.235,  0.34 , -0.065],\n",
+       "          [ 0.   , -0.02 , -0.02 , ..., -0.225,  0.34 , -0.065]],\n",
+       "         shape=(5000, 12)),\n",
+       "   {'fs': 500,\n",
+       "    'sig_len': 5000,\n",
+       "    'n_sig': 12,\n",
+       "    'base_date': datetime.date(2180, 8, 6),\n",
+       "    'base_time': datetime.time(9, 7),\n",
+       "    'units': ['mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV',\n",
+       "     'mV'],\n",
+       "    'sig_name': ['I',\n",
+       "     'II',\n",
+       "     'III',\n",
+       "     'aVR',\n",
+       "     'aVF',\n",
+       "     'aVL',\n",
+       "     'V1',\n",
+       "     'V2',\n",
+       "     'V3',\n",
+       "     'V4',\n",
+       "     'V5',\n",
+       "     'V6'],\n",
+       "    'comments': ['<subject_id>: 101']}]},\n",
+       " 'ehr': {np.int64(101):    subject_id  hadm_id         admittime  \\\n",
+       "  0         101        1  24/02/2196 14:38   \n",
+       "  1         101        2  17/09/2153 17:08   \n",
+       "  2         101        3  18/08/2134 02:02   \n",
+       "  \n",
+       "                                      first_careunit  \n",
+       "  0                                   Neuro Stepdown  \n",
+       "  1  Neuro Surgical Intensive Care Unit (Neuro SICU)  \n",
+       "  2                               Neuro Intermediate  },\n",
+       " 'echo': {np.int64(101): {'NumberOfFrames': '58',\n",
+       "   'Rows': 708,\n",
+       "   'Columns': 1016,\n",
+       "   'FrameTime': '33.6842',\n",
+       "   'CineRate': '30'}}}"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "ds[4]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "c0defa9b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from group_test.test_102 import *"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "04f7829a",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "   subject_id  hadm_id         admittime       first_careunit\n",
+      "0         102        4  13/11/2111 23:39  Trauma SICU (TSICU)\n",
+      "1         102        5  04/08/2113 18:46  Trauma SICU (TSICU)\n",
+      "2\n"
+     ]
+    }
+   ],
+   "source": [
+    "test_by_id(102)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "mmai-hackathon",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}