diff --git a/docs/experiments-arctic.md b/docs/experiments-arctic.md
new file mode 100644
index 000000000..1db71e011
--- /dev/null
+++ b/docs/experiments-arctic.md
@@ -0,0 +1,50 @@
+# Pyserini: Reproducing Arctic Results
+
+## MS marco v2.1 doc
+In order to handle msmarco v2.1 dataset's large size, we have the indexes divided in two partitions. Thus we need to perform retrieval runs for both of the indexes.
+
+```bash
+python -m pyserini.search.faiss --index /store/scratch/sjupadhy/indexes/msmarco-v2.1-snowflake-arctic-embed-l-1 \
+--topics /store/scratch/sjupadhy/msmarco-v2.1-snowflake-arctic-embed-l/topics/topics.msmarco-v2-doc.dev.json \ 
+--encoded-queries /store/scratch/sjupadhy/queries/msmarco-v2.1-dev-snowflake-arctic-embed-l/ \
+--output run.msmarco-v2.1-doc.arctic-embed-l-1.dev.txt \
+--hits 2000 --threads 16 --batch-size 128
+
+
+python -m pyserini.search.faiss --index /store/scratch/sjupadhy/indexes/msmarco-v2.1-snowflake-arctic-embed-l-2 \
+--topics /store/scratch/sjupadhy/msmarco-v2.1-snowflake-arctic-embed-l/topics/topics.msmarco-v2-doc.dev.json \ 
+--encoded-queries /store/scratch/sjupadhy/queries/msmarco-v2.1-dev-snowflake-arctic-embed-l/ \
+--output run.msmarco-v2.1-doc.arctic-embed-l-2.dev.txt \
+--hits 2000 --threads 16 --batch-size 128
+```
+
+### Merging and compiling docwise results
+As the available embeddings refer to doc segments, we need to complile doc wise results. Thus we merge and compile them with:
+```bash
+python scripts/arctic/merge_retrieved_results.py --arctic_run_folder arctic_runs \
+--output_file run.msmarco-v2.1-doc.arctic-embed-l-merged.dev.txt \
+--k 1000
+```
+
+### Evaluation
+```bash
+python -m pyserini.eval.trec_eval -c -m recall.1000 -m recall.100 -m ndcg_cut.10 /store/scratch/sjupadhy/msmarco-v2.1-snowflake-arctic-embed-l/qrels/qrels_qrels.msmarco-v2.1-doc.dev.txt run.msmarco-v2.1-doc.arctic-embed-l-merged.dev.txt
+Results:
+recall_1000           	all	0.9408
+recall_100            	all	0.8513
+ndcg_cut_10           	all	0.3583
+```
+
+## BEIR
+Retrieval run on NQ subdataset:
+```bash
+python -m pyserini.search.faiss --threads 16 --batch-size 512 --index beir-v1.0.0-nq.arctic-embed-m-v1.5 --topics beir-v1.0.0-nq-test --encoded-queries snowflake-arctic-embed-m-v1.5-beir-v1.0.0-nq-test  --output run.beir.arctic-embed.nq.txt --hits 1000
+```
+
+### Evaluation
+```bash
+python -m pyserini.eval.trec_eval   -c -m ndcg_cut.10  -m recall.1000 -m recall.100 beir-v1.0.0-nq-test   run.beir.arctic-embed.nq.txt
+Results:
+recall_1000             all     0.9951
+ndcg_cut_10             all     0.6244
+```
\ No newline at end of file
diff --git a/pyserini/encoded_query_info.py b/pyserini/encoded_query_info.py
index e3dabcb06..5fd84e757 100644
--- a/pyserini/encoded_query_info.py
+++ b/pyserini/encoded_query_info.py
@@ -1034,5 +1034,15 @@
         "size (bytes)": 135674,
         "total_queries": 49,
         "downloaded": False
+    },
+    "snowflake-arctic-embed-m-v1.5-beir-v1.0.0-nq-test": {
+        "description": "BEIR v1.0.0 nq test queries encoded by Snowflake arctic-embed-m-v1.5.",
+        "urls": [
+            "https://www.dropbox.com/scl/fi/i3du2elac5gxlvn595ioc/query-embedding-snowflake-arctic-embed-m-v1.5-beir-v1.0.0-nq-test-20240908.tar.gz?dl=1&rlkey=wkwfg8qpputewgixj7k7osxdv&st=zdlh9nrz"
+        ],
+        "md5": "9d869f5503c1b2606035ac0d512b666b",
+        "size (bytes)": 9461363,
+        "total_queries": 3452,
+        "downloaded": False
     },       
 }
diff --git a/pyserini/prebuilt_index_info.py b/pyserini/prebuilt_index_info.py
index b06964b5c..59697eaed 100644
--- a/pyserini/prebuilt_index_info.py
+++ b/pyserini/prebuilt_index_info.py
@@ -5159,7 +5159,20 @@ def add_lucene_index_info(enum, info, readme):
         "documents": 5183,
         "downloaded": False,
         "texts": "beir-v1.0.0-scifact.flat"
-    }
+    },
+    "beir-v1.0.0-nq.arctic-embed-m-v1.5": {
+        "description": "Faiss flat index for BEIR (v1.0.0): NQ, encoded by Arctic embed-m-v1.5.",
+        "filename": "faiss-flat.beir-v1.0.0-nq.arctic-embed-m-v1.5.20240908.tar.gz",
+        "readme": "faiss-flat.beir-v1.0.0.arctic-embed-m-v1.5.20240908.README.md",
+        "urls": [
+            "https://www.dropbox.com/scl/fi/abu8kn936rrxd85dk6sqy/faiss-flat.beir-v1.0.0-nq.arctic-embed-m-v1.5.20240908.tar.gz?dl=1&rlkey=9s7hy66d8en0nbv3ih6wjxh95&st=ooinw62c"
+        ],
+        "md5": "9a0dfded63e4554a002866ea57f4a30e",
+        "size compressed (bytes)": 7617697773,
+        "documents": 2681468,
+        "downloaded": False,
+        "texts": "beir-v1.0.0-nq.flat"
+    },
 }
 
 FAISS_INDEX_INFO_MRTYDI = {
diff --git a/pyserini/resources/index-metadata/faiss-flat.msmarco-v2.1-doc.arctic-embed-l.20240824.README.md b/pyserini/resources/index-metadata/faiss-flat.msmarco-v2.1-doc.arctic-embed-l.20240824.README.md
new file mode 100644
index 000000000..5eb4add3d
--- /dev/null
+++ b/pyserini/resources/index-metadata/faiss-flat.msmarco-v2.1-doc.arctic-embed-l.20240824.README.md
@@ -0,0 +1,26 @@
+# msmarco-v2.1-arctic-embed-l
+
+Faiss FlatIP indexes of msmarco v2.1 encoded by Snowflake embed-l. These indexes were generated on 2024/08/26 on `orca`.
+
+The indexes were generated from indexing embeddings available on [Huggingface](https://huggingface.co/datasets/Snowflake/msmarco-v2.1-snowflake-arctic-embed-l).
+
+## Preparation
+Due to msmarco v2.1 dataset's large size, indexes needed to be divided in two parts.
+
+```bash
+python scripts/arctic/convert_embeddings.py --embeddings_folder /store/scratch/sjupadhy/msmarco-v2.1-snowflake-arctic-embed-l/corpus \
+--output /store/scratch/sjupadhy/indexes/msmarco-v2.1-dev-snowflake-arctic-embed-l-1 \
+--indices 0_30
+
+python scripts/arctic/convert_embeddings.py --embeddings_folder /store/scratch/sjupadhy/msmarco-v2.1-snowflake-arctic-embed-l/corpus \
+--output /store/scratch/sjupadhy/indexes/msmarco-v2.1-dev-snowflake-arctic-embed-l-2 \
+--indices 30_59
+```
+
+### Topic embeddings
+```bash
+python scripts/arctic/convert_queries.py --embedding_path /store/scratch/sjupadhy/msmarco-v2.1-snowflake-arctic-embed-l/topics/snowflake-arctic-embed-l-topics.msmarco-v2-doc.dev.parquet \
+--output /store/scratch/sjupadhy/queries/msmarco-v2.1-dev-snowflake-arctic-embed-l
+
+```
+
diff --git a/pyserini/util.py b/pyserini/util.py
index 1649b3585..80ec5ba9f 100644
--- a/pyserini/util.py
+++ b/pyserini/util.py
@@ -70,6 +70,7 @@ def download_url(url, save_dir, local_filename=None, md5=None, force=False, verb
     if not local_filename:
         filename = url.split('/')[-1]
         filename = re.sub('\\?dl=1$', '', filename)  # Remove the Dropbox 'force download' parameter
+        filename = re.sub(r'\?.*$', '', filename)
     else:
         # Otherwise, use the specified local_filename:
         filename = local_filename
diff --git a/scripts/arctic/convert_embeddings.py b/scripts/arctic/convert_embeddings.py
new file mode 100644
index 000000000..fca3e3e41
--- /dev/null
+++ b/scripts/arctic/convert_embeddings.py
@@ -0,0 +1,74 @@
+"""Converts the hgf embeddings for documents into pyserini compatible format."""
+
+import argparse
+import os
+import pandas as pd
+import faiss
+import numpy as np
+from tqdm import tqdm
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--embeddings_folder",
+        type=str,
+        required=True,
+        help="Path to corpus embeddings folder downloaded from hgf.",
+    )
+    parser.add_argument(
+        "--output", type=str, help="Path to store faiss IndexFlatIP.", required=True
+    )
+    parser.add_argument(
+        "--indices",
+        type=str,
+        help="Start and end index seperated with _ or full.",
+        required=False,
+        default="full",
+    )
+    parser.add_argument(
+        "--start_filter",
+        type=str,
+        help="Filter to be used with string start.",
+        required=False,
+        default="",
+    )
+
+    args = parser.parse_args()
+
+    os.makedirs(args.output, exist_ok=True)
+
+    folder_path = args.embeddings_folder
+
+    files = [file for file in os.listdir(folder_path) if file.endswith(".parquet") and file.startswith(args.start_filter)]
+
+    if args.indices == "full":
+        start = 0
+        end = len(files)
+    else:
+        indices = args.indices.split("_")
+        start = int(indices[0])
+        end = int(indices[1])
+
+    all_embeddings = []
+    doc_ids = []
+    for file_name in tqdm(files[start:end]):
+        file_path = os.path.join(folder_path, file_name)
+        df = pd.read_parquet(file_path)
+        embeddings = df["VECTOR_MAIN"].tolist()
+        embeddings = np.array(embeddings)
+        dim = embeddings[0].shape[0]
+        faiss.normalize_L2(embeddings)
+        all_embeddings.append(embeddings.reshape(-1, dim))
+        doc_ids.extend(df["DOC_ID"].tolist())
+
+    combined_embeddings = np.vstack(all_embeddings)
+
+    index = faiss.IndexFlatIP(combined_embeddings.shape[1])
+    index.add(combined_embeddings)
+    faiss.write_index(index, os.path.join(args.output, "index"))
+
+    file_path = os.path.join(args.output, "docid")
+
+    with open(file_path, "w") as file:
+        for value in doc_ids:
+            file.write(f"{value}\n")
diff --git a/scripts/arctic/convert_topics.py b/scripts/arctic/convert_topics.py
new file mode 100644
index 000000000..50bc60774
--- /dev/null
+++ b/scripts/arctic/convert_topics.py
@@ -0,0 +1,47 @@
+"""Converts the hgf embeddings for topics into pyserini compatible format.
+
+python scripts/arctic/convert_queries.py --embedding_path /store/scratch/sjupadhy/msmarco-v2.1-snowflake-arctic-embed-l/topics/snowflake-arctic-embed-l-topics.msmarco-v2-doc.dev.parquet --output /store/scratch/sjupadhy/queries/msmarco-v2.1-dev-snowflake-arctic-embed-l
+"""
+
+import argparse
+import os
+
+import faiss
+import numpy as np
+import pandas as pd
+
+from pyserini.search import get_topics
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--embedding_path",
+        type=str,
+        required=True,
+        help="Path to corpus embeddings file for topics downloaded from hgf.",
+    )
+    parser.add_argument(
+        "--output", type=str, help="Path to store embedding.pkl.", required=True
+    )
+    parser.add_argument(
+        "--topic", type=str, help="Pyserini topic name.", required=True
+    )
+
+    args = parser.parse_args()
+
+    os.makedirs(args.output, exist_ok=True)
+
+    df = pd.read_parquet(args.embedding_path)
+    if "embedding" not in df.columns:
+        df.rename(columns={"VECTOR_MAIN": "embedding", "QUERY_ID": "id"}, inplace=True)
+    array_2d = np.vstack(df["embedding"].values)
+    faiss.normalize_L2(array_2d)
+    df["embedding"] = [array_2d[i, :] for i in range(array_2d.shape[0])]
+
+    if "text" not in df.columns:
+        topics_mapping = get_topics(args.topic)
+        text_list = [topics_mapping.get(topic).get("title") for topic in df["id"].to_list()]
+        df["text"] = text_list
+
+    df.to_pickle(os.path.join(args.output, "embedding.pkl"))
diff --git a/scripts/arctic/merge_retrieved_results.py b/scripts/arctic/merge_retrieved_results.py
new file mode 100644
index 000000000..9cde37a10
--- /dev/null
+++ b/scripts/arctic/merge_retrieved_results.py
@@ -0,0 +1,66 @@
+import argparse
+import os
+from typing import List
+
+
+def merge_retrieved(shard_files: List[str], output_file: str, top_n: int) -> None:
+    merged_results = {}
+    for shard_file in shard_files:
+        print(f"Loading shard {shard_file} ")
+        with open(shard_file, "r") as f:
+            for line in f:
+                data = line.split()
+                if data[0] not in merged_results:
+                    merged_results[data[0]] = []
+                merged_results[data[0]].append((data[2], data[4]))
+    print("Shards all loaded, merging results and sorting by score")
+    run = {}
+
+    for query_id, doc_scores in merged_results.items():
+        doc_score_dict = {}
+        for passage_id, score in doc_scores:
+            doc_id = passage_id.split("#")[0]
+            if doc_id not in doc_score_dict:
+                doc_score_dict[doc_id] = (
+                    -1
+                )  # scores are in range -1 to 1 on similairty so starting at -1 is floor
+            if float(score) > float(doc_score_dict[doc_id]):
+                doc_score_dict[doc_id] = score
+        top_docs = sorted(doc_score_dict.items(), key=lambda x: x[1], reverse=True)[
+            :top_n
+        ]
+        run[query_id] = {
+            doc_id: round(float(score) * 100, 2) for doc_id, score in top_docs
+        }
+    results = []
+    for qid in run:
+        for index, doc_id in enumerate(run[qid]):
+            results.append(
+                f"{qid} Q0 {doc_id} {index + 1} {run[qid][doc_id]} faiss-merged"
+            )
+
+    with open(output_file, "w") as f:
+        for line in results:
+            f.write(f"{line}\n")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--arctic_run_folder",
+        type=str,
+        required=True,
+        help="Path to run files that needs to be combined.",
+    )
+    parser.add_argument(
+        "--output_file", type=str, help="Path to docwise merged file", required=True
+    )
+    parser.add_argument("--k", default=1000, help="k value", type=int)
+    args = parser.parse_args()
+
+    files = [
+        os.path.join(args.arctic_run_folder, file)
+        for file in os.listdir(args.arctic_run_folder)
+    ]
+
+    merge_retrieved(files, args.output_file, args.k)