diff --git a/python/perf-kernels/tools/rocm-triton-prof/README.md b/python/perf-kernels/tools/rocm-triton-prof/README.md
index 350c0aa0ce00..04b9bf5c2cfb 100644
--- a/python/perf-kernels/tools/rocm-triton-prof/README.md
+++ b/python/perf-kernels/tools/rocm-triton-prof/README.md
@@ -129,7 +129,20 @@ max      261.981219
 dtype: float64
 ```
 
+You can also run the tool using a yaml configuration file as well. For example:
+
+```bash
+$ cat ./config.yaml
+kernel: ".*fwd"
+cmd: 'python3 ./flash-attention.py -b 2 -hq 4 -hk 4 -sq 113 -sk 123 -d 1 -causal -layout bhsd'
+
+$ python3 ./rocm-triton-prof.py -f ./config.yaml
+```
+
+Note, you can use a regular expression for the kernel name (see example above). The tool is
+going to abort of more than one kernel match the given regular expression.
+
 ### Known limits
 
-The tool currently supports only FP64, FP32 and FP16 operations.
-Note, it can be extended to supoprt other data types.
+The tool currently supports only FP64, FP32 and FP16 operations. Note, it can be extended
+to supoprt other data types.
diff --git a/python/perf-kernels/tools/rocm-triton-prof/rocm-triton-prof.py b/python/perf-kernels/tools/rocm-triton-prof/rocm-triton-prof.py
index feafb74db311..d375fa1309e2 100755
--- a/python/perf-kernels/tools/rocm-triton-prof/rocm-triton-prof.py
+++ b/python/perf-kernels/tools/rocm-triton-prof/rocm-triton-prof.py
@@ -104,9 +104,22 @@ def filter(df, name):
     return df[df['Kernel_Name'] == name]
 
 
-def process_files(metrics_dir, timing_dir, kernel_name, verbose):
+def find_kernel_name(df, kernel_expr, verbose):
+    names = df['Kernel_Name']
+    raw_kernel_expr = r'{}'.format(kernel_expr)
+    names_set = set(names[names.str.match(raw_kernel_expr)])
+    if len(names_set) != 1:
+        raise RuntimeError(f'Error: found several kernels matching `kernel` regex i.e., {names_set}')
+    kernel_name = next(iter(names_set))
+    if verbose:
+        print(f'\nTracking: `{kernel_name}` kernel\n')
+    return next(iter(names_set))
+
+
+def process_files(metrics_dir, timing_dir, kernel_expr, verbose):
     timing_file = find_file(timing_dir, re.compile(r'.*kernel_trace.csv'))
     df = pd.read_csv(timing_file)
+    kernel_name = find_kernel_name(df, kernel_expr, verbose)
     df = filter(df, kernel_name)
     timing = df['End_Timestamp'] - df['Start_Timestamp']
     print('Timing info in `nsec`:')
@@ -247,10 +260,33 @@ def main(args):
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
-    parser.add_argument("-k", "--kernel", type=str, required=True, help="name of a kernel")
-    parser.add_argument('-c', '--cmd', required=True, nargs=argparse.REMAINDER, help='user command')
+    parser.add_argument("-f", "--file", type=str, required=False, help="config file")
+    parser.add_argument("-k", "--kernel", type=str, required=False, help="name of a kernel")
+    parser.add_argument('-c', '--cmd', required=False, nargs=argparse.REMAINDER, help='user command')
     parser.add_argument("--display-only", action='store_true', help='display info without running')
     parser.add_argument("-v", "--verbose", action='store_true', help='verbose output')
     args = parser.parse_args()
-
+    if args.file:
+        if os.path.exists(args.file):
+            RuntimeError(f'file `{args.file}` does not exist')
+        with open(args.file, 'r') as stream:
+            data = yaml.safe_load(stream)
+            args.kernel = data['kernel']
+
+            args.cmd = data['cmd']
+            assert (type(args.cmd) is str)
+            args.cmd = args.cmd.split(' ')
+
+            args.display_only = args.display_only
+            if 'display_only' in data:
+                args.display_only = data['display_only']
+
+            args.verbose = args.verbose
+            if 'verbose' in data:
+                args.verbose = data['verbose']
+    else:
+        if not hasattr(args, 'kernel'):
+            RuntimeError('kernel expr must be specified (--kernel)')
+        if not hasattr(args, 'cmd'):
+            RuntimeError('program command must be specified (--cmd)')
     main(args)