diff --git a/docs/abstractions2.py b/docs/abstractions2.py
index 7112efe4ff..3e5deafab0 100644
--- a/docs/abstractions2.py
+++ b/docs/abstractions2.py
@@ -73,7 +73,7 @@ assert out.as_buffer().cast('I')[0] == 5
 print("******** third, the LazyBuffer ***********")
 
 from tinygrad.lazy import LazyBuffer, LoadOps
-from tinygrad.realize import run_schedule, create_schedule
+from tinygrad.engine.realize import run_schedule, create_schedule
 
 # allocate some values + load in values
 a = LazyBuffer.loadop(LoadOps.EMPTY, (1,), dtypes.int32, DEVICE)
diff --git a/examples/benchmark_train_efficientnet.py b/examples/benchmark_train_efficientnet.py
index 6e96d8762e..fb26e2893a 100755
--- a/examples/benchmark_train_efficientnet.py
+++ b/examples/benchmark_train_efficientnet.py
@@ -7,7 +7,7 @@ from tinygrad.nn.state import get_parameters
 from tinygrad.nn import optim
 from tinygrad import Tensor, GlobalCounters
 from tinygrad.helpers import getenv
-from tinygrad.features.jit import CacheCollector
+from tinygrad.engine.jit import CacheCollector
 
 def tensors_allocated():
   return sum(isinstance(x, Tensor) for x in gc.get_objects())
diff --git a/examples/efficientnet.py b/examples/efficientnet.py
index 7bfacd4d8c..863af3d124 100644
--- a/examples/efficientnet.py
+++ b/examples/efficientnet.py
@@ -9,7 +9,7 @@ import numpy as np
 from PIL import Image
 from tinygrad.tensor import Tensor
 from tinygrad.helpers import getenv, fetch, Timing
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from extra.models.efficientnet import EfficientNet
 np.set_printoptions(suppress=True)
 
diff --git a/examples/handcode_resnet50_opt.py b/examples/handcode_resnet50_opt.py
index ee2acba2f3..4ce146c2ab 100644
--- a/examples/handcode_resnet50_opt.py
+++ b/examples/handcode_resnet50_opt.py
@@ -8,7 +8,7 @@ from tinygrad.features.search import time_linearizer, beam_search, bufs_from_lin
 from tinygrad.helpers import ansilen, DEBUG, getenv
 from tinygrad.shape.symbolic import sym_infer
 from tinygrad.dtype import dtypes
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 
 if __name__ == "__main__":
   if getenv("HALF"):
diff --git a/examples/mlperf/model_eval.py b/examples/mlperf/model_eval.py
index 7dadc9ef36..69d5be853e 100644
--- a/examples/mlperf/model_eval.py
+++ b/examples/mlperf/model_eval.py
@@ -3,7 +3,7 @@ start = time.perf_counter()
 from pathlib import Path
 import numpy as np
 from tinygrad import Tensor, Device, dtypes, GlobalCounters
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad.nn.state import get_parameters, load_state_dict, safe_load
 from tinygrad.helpers import getenv, Timing
 from examples.mlperf import helpers
@@ -103,7 +103,7 @@ def eval_retinanet():
   coco_eval = COCOeval(coco, iouType="bbox")
   coco_evalimgs, evaluated_imgs, ncats, narea = [], [], len(coco_eval.params.catIds), len(coco_eval.params.areaRng)
 
-  from tinygrad.features.jit import TinyJit
+  from tinygrad.engine.jit import TinyJit
   mdlrun = TinyJit(lambda x: mdl(input_fixup(x)).realize())
 
   n, bs = 0, 8
diff --git a/examples/stable_diffusion.py b/examples/stable_diffusion.py
index 96b9c744ed..c7048c0a21 100644
--- a/examples/stable_diffusion.py
+++ b/examples/stable_diffusion.py
@@ -13,7 +13,7 @@ from tinygrad import Device, GlobalCounters, dtypes, Tensor
 from tinygrad.helpers import Timing, Context, getenv, fetch, colored
 from tinygrad.nn import Conv2d, Linear, GroupNorm, LayerNorm, Embedding
 from tinygrad.nn.state import torch_load, load_state_dict, get_state_dict
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 
 class AttnBlock:
   def __init__(self, in_channels):
diff --git a/examples/vits.py b/examples/vits.py
index 37cb61e0c3..1708b3568f 100644
--- a/examples/vits.py
+++ b/examples/vits.py
@@ -9,7 +9,7 @@ from tinygrad import nn, dtypes
 from tinygrad.helpers import fetch
 from tinygrad.nn.state import torch_load
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from unidecode import unidecode
 
 LRELU_SLOPE = 0.1
diff --git a/examples/whisper.py b/examples/whisper.py
index 536f2134bd..5caf1f2f0c 100644
--- a/examples/whisper.py
+++ b/examples/whisper.py
@@ -6,7 +6,7 @@ import base64
 import multiprocessing
 import numpy as np
 from typing import Optional, Union, Literal, List
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad.nn.state import torch_load, load_state_dict
 from tinygrad.helpers import getenv, DEBUG, CI, fetch
 import tinygrad.nn as nn
diff --git a/extra/autopad.py b/extra/autopad.py
index 7aebc2358f..25f739682f 100644
--- a/extra/autopad.py
+++ b/extra/autopad.py
@@ -3,7 +3,7 @@ from tinygrad.ops import LoadOps
 from tinygrad.codegen.linearizer import Linearizer
 from test.external.fuzz_linearizer import run_linearizer
 from tinygrad.codegen.kernel import Opt, OptOps
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 
 N = 17**3
 
diff --git a/extra/datasets/kits19.py b/extra/datasets/kits19.py
index 12bcd3cfa6..ef32a936bf 100644
--- a/extra/datasets/kits19.py
+++ b/extra/datasets/kits19.py
@@ -91,7 +91,7 @@ def pad_input(volume, roi_shape, strides, padding_mode="constant", padding_val=-
   return F.pad(torch.from_numpy(volume), paddings, mode=padding_mode, value=padding_val).numpy(), paddings
 
 def sliding_window_inference(model, inputs, labels, roi_shape=(128, 128, 128), overlap=0.5):
-  from tinygrad.features.jit import TinyJit
+  from tinygrad.engine.jit import TinyJit
   mdl_run = TinyJit(lambda x: model(x).realize())
   image_shape, dim = list(inputs.shape[2:]), len(inputs.shape[2:])
   strides = [int(roi_shape[i] * (1 - overlap)) for i in range(dim)]
diff --git a/extra/export_model.py b/extra/export_model.py
index 9c931f4a13..fd7cba5ae3 100644
--- a/extra/export_model.py
+++ b/extra/export_model.py
@@ -1,7 +1,7 @@
 from typing import Tuple, Dict, List
 from tinygrad.dtype import DType
 from tinygrad.tensor import Device, Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad.nn.state import get_state_dict
 from tinygrad.dtype import dtypes
 import json
diff --git a/extra/gemm/metal_conv.py b/extra/gemm/metal_conv.py
index 0d0c1f4ac7..1dc7a18ea2 100644
--- a/extra/gemm/metal_conv.py
+++ b/extra/gemm/metal_conv.py
@@ -32,7 +32,7 @@ except RuntimeError:
   print("no torch metal conv")
 
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad import Device
 b = Tensor(nb)
 c = Tensor(nc)
diff --git a/extra/gemm/metal_matmul.py b/extra/gemm/metal_matmul.py
index 255eedf24a..76cd1473b3 100644
--- a/extra/gemm/metal_matmul.py
+++ b/extra/gemm/metal_matmul.py
@@ -116,7 +116,7 @@ tm = min([torch_prog(b, c) for _ in range(20)])
 print(f"{N*N:10d} {tm*1e6:9.2f} us, would be {FLOPS*1e-9/tm:9.2f} GFLOPS matmul in torch")
 
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 b = Tensor(nb)
 c = Tensor(nc)
 # TODO: slowness without the JIT I suspect comes from a lack of a caching allocator
diff --git a/extra/gemm/metal_matvec.py b/extra/gemm/metal_matvec.py
index ce9b34de33..0317312ec5 100644
--- a/extra/gemm/metal_matvec.py
+++ b/extra/gemm/metal_matvec.py
@@ -4,7 +4,7 @@ import numpy as np
 import time, torch, torch.mps
 
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad import Device, GlobalCounters, dtypes
 from tinygrad.helpers import colored, getenv, CI, flat_mv
 
@@ -108,7 +108,7 @@ metalalloc.copyout(flat_mv(metal_a.data), a)
 np.testing.assert_allclose(metal_a, torch_a, atol=5e-3)
 
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 b = Tensor(nb)
 c = Tensor(nc)
 # TODO: slowness without the JIT I suspect comes from a lack of a caching allocator
diff --git a/extra/gemm/tvm_gemm.py b/extra/gemm/tvm_gemm.py
index d89d189852..1c5253b604 100644
--- a/extra/gemm/tvm_gemm.py
+++ b/extra/gemm/tvm_gemm.py
@@ -30,7 +30,7 @@ except ImportError:
 
 import os
 from tinygrad.tensor import Tensor
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 
 # define the compute
 A = Tensor.rand(M, K, device="clang")
diff --git a/extra/models/rnnt.py b/extra/models/rnnt.py
index d7082c4be6..3902fe2313 100644
--- a/extra/models/rnnt.py
+++ b/extra/models/rnnt.py
@@ -1,5 +1,5 @@
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad.nn import Linear, Embedding
 from tinygrad.helpers import fetch
 import numpy as np
diff --git a/extra/training.py b/extra/training.py
index 7cd353830b..9145830840 100644
--- a/extra/training.py
+++ b/extra/training.py
@@ -2,7 +2,7 @@ import numpy as np
 from tqdm import trange
 from tinygrad.tensor import Tensor
 from tinygrad.helpers import CI
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 
 
 def train(model, X_train, Y_train, optim, steps, BS=128, lossfn=lambda out,y: out.sparse_categorical_crossentropy(y),
diff --git a/openpilot/compile2.py b/openpilot/compile2.py
index acedf9b0fc..e82119bae4 100644
--- a/openpilot/compile2.py
+++ b/openpilot/compile2.py
@@ -16,7 +16,7 @@ from extra.onnx import get_run_onnx
 from tinygrad import Tensor, Device, GlobalCounters, dtypes
 from tinygrad.dtype import ImageDType
 from tinygrad.helpers import partition, Context, fetch, getenv, GRAPH, DEBUG
-from tinygrad.realize import run_schedule, lower_schedule_item, create_schedule
+from tinygrad.engine.realize import run_schedule, lower_schedule_item, create_schedule
 from tinygrad.ops import LoadOps, ScheduleItem
 Device.DEFAULT = "GPU"
 
diff --git a/setup.py b/setup.py
index d0446d80c0..e69ea3b572 100644
--- a/setup.py
+++ b/setup.py
@@ -14,7 +14,7 @@ setup(name='tinygrad',
       license='MIT',
       long_description=long_description,
       long_description_content_type='text/markdown',
-      packages = ['tinygrad', 'tinygrad.runtime.autogen', 'tinygrad.codegen', 'tinygrad.nn', 'tinygrad.renderer',
+      packages = ['tinygrad', 'tinygrad.runtime.autogen', 'tinygrad.codegen', 'tinygrad.nn', 'tinygrad.renderer', 'tinygrad.engine',
                   'tinygrad.runtime', 'tinygrad.runtime.driver', 'tinygrad.runtime.graph', 'tinygrad.shape', 'tinygrad.features'],
       classifiers=[
         "Programming Language :: Python :: 3",
@@ -54,7 +54,7 @@ setup(name='tinygrad',
             "hypothesis",
         ],
         'testing_tf': [
-            "tensorflow",
+            "tensorflow==2.15.1",
             "tensorflow_addons",
         ]
       },
diff --git a/test/external/external_benchmark_multitensor_allreduce.py b/test/external/external_benchmark_multitensor_allreduce.py
index 5a097628d6..24496fe3c3 100644
--- a/test/external/external_benchmark_multitensor_allreduce.py
+++ b/test/external/external_benchmark_multitensor_allreduce.py
@@ -3,8 +3,8 @@ from tinygrad import Tensor, Device
 from tinygrad.lazy import LazyBuffer
 from tinygrad.ops import ReduceOps, GlobalCounters
 from tinygrad.features.multi import MultiLazyBuffer, all_reduce
-from tinygrad.features.jit import TinyJit
-from tinygrad.realize import create_schedule, run_schedule
+from tinygrad.engine.jit import TinyJit
+from tinygrad.engine.realize import create_schedule, run_schedule
 from tinygrad.helpers import getenv, Context, RING
 from typing import List, Union
 
diff --git a/test/external/external_model_benchmark.py b/test/external/external_model_benchmark.py
index db4fad333b..57deec65b6 100644
--- a/test/external/external_model_benchmark.py
+++ b/test/external/external_model_benchmark.py
@@ -66,7 +66,7 @@ def benchmark_model(m, devices, validate_outs=False):
     tinygrad_model = get_run_onnx(onnx_model)
     benchmark(m, f"tinygrad_{device.lower()}_jitless", lambda: {k:v.numpy() for k,v in tinygrad_model(inputs).items()})
 
-    from tinygrad.features.jit import TinyJit
+    from tinygrad.engine.jit import TinyJit
     tinygrad_jitted_model = TinyJit(lambda **kwargs: {k:v.realize() for k,v in tinygrad_model(kwargs).items()})
     for _ in range(3): {k:v.numpy() for k,v in tinygrad_jitted_model(**inputs).items()}
     benchmark(m, f"tinygrad_{device.lower()}_jit", lambda: {k:v.numpy() for k,v in tinygrad_jitted_model(**inputs).items()}) # noqa: F821
diff --git a/test/external/external_test_hip_compile.py b/test/external/external_test_hip_compile.py
index e351d3a488..68fa2b7e3a 100644
--- a/test/external/external_test_hip_compile.py
+++ b/test/external/external_test_hip_compile.py
@@ -2,7 +2,7 @@ import time, unittest
 from tinygrad.runtime.driver.hip_comgr import compile_hip
 from tinygrad import Tensor
 from tinygrad.device import Device
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 from tinygrad.codegen.linearizer import Linearizer
 
 class TestHIPCompileSpeed(unittest.TestCase):
diff --git a/test/external/external_test_hsa_driver.py b/test/external/external_test_hsa_driver.py
index e108c01bd3..4dd8cded33 100644
--- a/test/external/external_test_hsa_driver.py
+++ b/test/external/external_test_hsa_driver.py
@@ -4,7 +4,7 @@ from tinygrad.device import Device, Buffer, BufferXfer
 from tinygrad.dtype import dtypes
 from tinygrad.runtime.driver.hsa import AQLQueue
 from tinygrad.runtime.graph.hsa import VirtAQLQueue, HSAGraph
-from tinygrad.features.jit import JitItem
+from tinygrad.engine.jit import JitItem
 
 def get_hsa_inc_prog(dev, inc=1):
   prg = f"""
diff --git a/test/external/external_test_jit_on_models.py b/test/external/external_test_jit_on_models.py
index 4ce5adfea4..92b18bdb03 100644
--- a/test/external/external_test_jit_on_models.py
+++ b/test/external/external_test_jit_on_models.py
@@ -2,7 +2,7 @@
 import unittest
 import numpy as np
 from tinygrad import Tensor, dtypes
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad.helpers import CI
 from test.helpers import derandomize_model
 
diff --git a/test/external/external_test_opt.py b/test/external/external_test_opt.py
index fdb349020f..75bacc273c 100644
--- a/test/external/external_test_opt.py
+++ b/test/external/external_test_opt.py
@@ -17,7 +17,7 @@ from tinygrad.helpers import getenv
 from tinygrad.nn import optim
 #from tinygrad.lazy import PUSH_PERMUTES
 PUSH_PERMUTES = False
-from tinygrad.features.jit import CacheCollector
+from tinygrad.engine.jit import CacheCollector
 
 class CLCache:
   def __init__(self, allowed=None, strict=False, preclear=True, var_vals=None):
diff --git a/test/external/external_test_uops_graphing.py b/test/external/external_test_uops_graphing.py
index 56bd18cd7b..82bdd0b1ac 100644
--- a/test/external/external_test_uops_graphing.py
+++ b/test/external/external_test_uops_graphing.py
@@ -4,7 +4,7 @@ from tinygrad.tensor import Tensor
 from tinygrad.codegen.linearizer import Linearizer
 from tinygrad.renderer.cstyle import OpenCLRenderer
 from tinygrad.features.graph import graph_uops
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 from tinygrad.nn import Conv2d
 
 class TestUopsGraph(unittest.TestCase):
diff --git a/test/models/test_real_world.py b/test/models/test_real_world.py
index 4615006910..aee49587df 100644
--- a/test/models/test_real_world.py
+++ b/test/models/test_real_world.py
@@ -2,7 +2,7 @@ import unittest, time, gc
 import numpy as np
 from tinygrad.nn import optim
 from tinygrad.nn.state import get_parameters
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad import Tensor, Device, GlobalCounters, dtypes
 from tinygrad.helpers import CI, getenv
 from tinygrad.shape.symbolic import Variable
diff --git a/test/test_conv_shapetracker.py b/test/test_conv_shapetracker.py
index 5d46be6335..acd0bda00f 100644
--- a/test/test_conv_shapetracker.py
+++ b/test/test_conv_shapetracker.py
@@ -3,7 +3,7 @@ import unittest
 from tinygrad.tensor import Tensor
 from tinygrad.ops import LoadOps
 from tinygrad.nn import Conv2d
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 
 class TestConvShapetracker(unittest.TestCase):
   def test_conv_3x3_one_view(self):
diff --git a/test/test_custom_function.py b/test/test_custom_function.py
index b376889f52..76d8c0a038 100644
--- a/test/test_custom_function.py
+++ b/test/test_custom_function.py
@@ -87,7 +87,7 @@ class TestCustomFunction(unittest.TestCase):
   @unittest.skipIf(Device.DEFAULT in ["CPU"], "atan2_cpu not jittable")
   def test_atan2_jit(self):
     # custom ops even work in the JIT!
-    from tinygrad.features.jit import TinyJit
+    from tinygrad.engine.jit import TinyJit
 
     @TinyJit
     def jitted_atan2(a:Tensor, b:Tensor) -> Tensor:
diff --git a/test/test_dtype_alu.py b/test/test_dtype_alu.py
index 93e581aeee..c141194572 100644
--- a/test/test_dtype_alu.py
+++ b/test/test_dtype_alu.py
@@ -6,7 +6,7 @@ import numpy as np
 from hypothesis import given, strategies as strat, settings
 from tinygrad.dtype import DType
 from tinygrad.helpers import CI, getenv
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 from tinygrad.ops import UnaryOps, get_lazyop_info
 from test.helpers import is_dtype_supported
 
diff --git a/test/test_fusion_op.py b/test/test_fusion_op.py
index b0c325ff95..5f73a500ee 100644
--- a/test/test_fusion_op.py
+++ b/test/test_fusion_op.py
@@ -2,7 +2,7 @@ import unittest
 import time
 import numpy as np
 from tinygrad import Tensor, dtypes
-from tinygrad.realize import run_schedule, create_schedule, lower_schedule_item
+from tinygrad.engine.realize import run_schedule, create_schedule, lower_schedule_item
 
 class TestFusionOp(unittest.TestCase):
   def test_contiguous_add(self):
diff --git a/test/test_jit.py b/test/test_jit.py
index 1c02aeada3..e9ca4b68b2 100644
--- a/test/test_jit.py
+++ b/test/test_jit.py
@@ -4,7 +4,7 @@ import numpy as np
 
 from test.helpers import assert_jit_cache_len
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad.device import Device
 from tinygrad.helpers import CI
 
diff --git a/test/test_lazybuffer.py b/test/test_lazybuffer.py
index 5217244faa..b0bb4b873e 100644
--- a/test/test_lazybuffer.py
+++ b/test/test_lazybuffer.py
@@ -3,7 +3,7 @@ import numpy as np
 import unittest
 from tinygrad import Tensor, Device, dtypes
 from tinygrad.lazy import LazyBuffer, ReduceOps
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 
 class TestLazyBuffer(unittest.TestCase):
   def test_fromcpu_shape_tracker(self):
diff --git a/test/test_lazyop.py b/test/test_lazyop.py
index da5d8a1227..34e707b5d6 100644
--- a/test/test_lazyop.py
+++ b/test/test_lazyop.py
@@ -1,6 +1,6 @@
 import unittest
 from tinygrad.tensor import Tensor
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 
 # stuff needed to unpack a kernel
 # ruff: noqa: F401
diff --git a/test/test_linearizer.py b/test/test_linearizer.py
index e9e6e88bbf..eab1b6ba17 100644
--- a/test/test_linearizer.py
+++ b/test/test_linearizer.py
@@ -9,8 +9,8 @@ from tinygrad.shape.shapetracker import ShapeTracker
 from tinygrad.shape.view import View
 from tinygrad.shape.symbolic import MulNode, Variable, NumNode, Node
 from tinygrad.tensor import Tensor
-from tinygrad.features.jit import CacheCollector
-from tinygrad.realize import create_schedule, run_schedule
+from tinygrad.engine.jit import CacheCollector
+from tinygrad.engine.realize import create_schedule, run_schedule
 from tinygrad.helpers import prod, Context
 from tinygrad.dtype import DType, dtypes
 from tinygrad.codegen.uops import UOpGraph
diff --git a/test/test_multitensor.py b/test/test_multitensor.py
index 6df21a6c4e..347d45ea1c 100644
--- a/test/test_multitensor.py
+++ b/test/test_multitensor.py
@@ -5,7 +5,7 @@ from tinygrad.device import BufferCopy
 from tinygrad.ops import LoadOps, ReduceOps
 from tinygrad.helpers import CI, prod, Context
 from tinygrad.nn.state import get_parameters, get_state_dict
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 from tinygrad.features.multi import all_reduce, MultiLazyBuffer
 from random import randint
 import numpy as np
diff --git a/test/test_schedule.py b/test/test_schedule.py
index 609a6db82a..e2128dd8db 100644
--- a/test/test_schedule.py
+++ b/test/test_schedule.py
@@ -9,7 +9,7 @@ from tinygrad.ops import LoadOps
 from tinygrad.helpers import DEBUG, GRAPH
 from tinygrad.codegen.linearizer import Linearizer
 from tinygrad.features.graph import print_tree, realized_lazybuffer
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 from tinygrad import nn, dtypes
 
 def check_schedule(t:Tensor, allowed:int, to_prerealize:Optional[List[Tensor]]=None, filter_loadops=True):
diff --git a/test/test_search.py b/test/test_search.py
index 81a7100c61..987f156c0e 100644
--- a/test/test_search.py
+++ b/test/test_search.py
@@ -1,7 +1,7 @@
 import unittest
 
 from tinygrad.codegen.linearizer import Linearizer
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 from tinygrad.features.search import time_linearizer, bufs_from_lin
 from tinygrad.device import Device, Buffer
 from tinygrad.ops import LoadOps
diff --git a/test/test_speed_v_torch.py b/test/test_speed_v_torch.py
index 6b1e31fd34..01c2a599db 100644
--- a/test/test_speed_v_torch.py
+++ b/test/test_speed_v_torch.py
@@ -13,7 +13,7 @@ from tinygrad import Device, GlobalCounters
 from tinygrad.tensor import Tensor
 from tinygrad.nn import Conv2d
 from tinygrad.helpers import colored, getenv, CI
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 
 IN_CHANS = [int(x) for x in getenv("IN_CHANS", "4,16,64").split(",")]
 
diff --git a/test/test_symbolic_jit.py b/test/test_symbolic_jit.py
index 96c61f5f2b..be8e854f2f 100644
--- a/test/test_symbolic_jit.py
+++ b/test/test_symbolic_jit.py
@@ -1,7 +1,7 @@
 import unittest
 
 from test.helpers import assert_jit_cache_len
-from tinygrad.features.jit import TinyJit
+from tinygrad.engine.jit import TinyJit
 from tinygrad.shape.symbolic import Variable
 from tinygrad.tensor import Tensor
 import numpy as np
diff --git a/test/test_uops.py b/test/test_uops.py
index 1f18a5ba27..5883cc206a 100644
--- a/test/test_uops.py
+++ b/test/test_uops.py
@@ -5,7 +5,7 @@ from tinygrad.tensor import Tensor
 from tinygrad.dtype import dtypes, DType, PtrDType
 from tinygrad.device import Buffer, Device, CompiledASTRunner
 from tinygrad.ops import UnaryOps, BinaryOps, TernaryOps
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 from tinygrad.codegen.linearizer import UOps, UOp
 from tinygrad.codegen.uops import exec_alu, UOpGraph
 from test.helpers import is_dtype_supported
diff --git a/test/test_uops_stats.py b/test/test_uops_stats.py
index 830f536ee3..6fce3c5ec4 100644
--- a/test/test_uops_stats.py
+++ b/test/test_uops_stats.py
@@ -1,6 +1,6 @@
 import unittest
 from tinygrad import Tensor
-from tinygrad.realize import create_schedule, lower_schedule_item
+from tinygrad.engine.realize import create_schedule, lower_schedule_item
 
 # TODO: can copy this in here when we remove it
 #from tinygrad.ops import get_lazyop_info
diff --git a/test/test_winograd.py b/test/test_winograd.py
index 29e7ef8ef2..8d00a3967d 100644
--- a/test/test_winograd.py
+++ b/test/test_winograd.py
@@ -3,7 +3,7 @@ from tinygrad import Tensor, GlobalCounters
 from tinygrad.helpers import Timing, CI, Profiling, WINO, DEBUG
 from tinygrad.ops import LoadOps
 from tinygrad.codegen.linearizer import Linearizer
-from tinygrad.realize import create_schedule
+from tinygrad.engine.realize import create_schedule
 
 class TestWinograd(unittest.TestCase):
   def setUp(self):
diff --git a/tinygrad/__init__.py b/tinygrad/__init__.py
index 3ffb0a955c..3c96deba96 100644
--- a/tinygrad/__init__.py
+++ b/tinygrad/__init__.py
@@ -1,5 +1,5 @@
 from tinygrad.tensor import Tensor            # noqa: F401
-from tinygrad.features.jit import TinyJit     # noqa: F401
+from tinygrad.engine.jit import TinyJit       # noqa: F401
 from tinygrad.shape.symbolic import Variable  # noqa: F401
 from tinygrad.dtype import dtypes             # noqa: F401
 from tinygrad.ops import GlobalCounters       # noqa: F401
diff --git a/tinygrad/device.py b/tinygrad/device.py
index 2bb0e0f324..96cf473d53 100644
--- a/tinygrad/device.py
+++ b/tinygrad/device.py
@@ -46,7 +46,7 @@ class JITRunner:
     self.mem_estimate:sint = 0
   def exec(self, rawbufs:List[Buffer], var_vals:Optional[Dict[Variable, int]]=None) -> Optional[float]:
     var_vals = var_vals if var_vals is not None else {}
-    from tinygrad.features.jit import CacheCollector
+    from tinygrad.engine.jit import CacheCollector
     et = self(rawbufs, var_vals)
     if CACHECOLLECTING: CacheCollector.add(self, rawbufs, var_vals)
     return et
diff --git a/tinygrad/features/jit.py b/tinygrad/engine/jit.py
similarity index 100%
rename from tinygrad/features/jit.py
rename to tinygrad/engine/jit.py
diff --git a/tinygrad/realize.py b/tinygrad/engine/realize.py
similarity index 100%
rename from tinygrad/realize.py
rename to tinygrad/engine/realize.py
diff --git a/tinygrad/runtime/graph/cuda.py b/tinygrad/runtime/graph/cuda.py
index 3fab54513a..ae30f81c8f 100644
--- a/tinygrad/runtime/graph/cuda.py
+++ b/tinygrad/runtime/graph/cuda.py
@@ -5,7 +5,7 @@ from tinygrad.helpers import init_c_var, GraphException
 from tinygrad.device import CompiledASTRunner, update_stats, Buffer, MultiDeviceJITGraph, BufferXfer
 from tinygrad.runtime.ops_cuda import CUDADevice, check, encode_args, cu_time_execution
 from tinygrad.shape.symbolic import Variable
-from tinygrad.features.jit import JitItem, get_input_replace, get_jit_stats, \
+from tinygrad.engine.jit import JitItem, get_input_replace, get_jit_stats, \
                                   get_jc_idxs_with_updatable_launch_dims, get_jc_idxs_with_updatable_var_vals
 
 class CUDAGraph(MultiDeviceJITGraph):
diff --git a/tinygrad/runtime/graph/hsa.py b/tinygrad/runtime/graph/hsa.py
index 763b060505..73c087598a 100644
--- a/tinygrad/runtime/graph/hsa.py
+++ b/tinygrad/runtime/graph/hsa.py
@@ -4,7 +4,7 @@ from tinygrad.helpers import GraphException, init_c_var, round_up
 from tinygrad.device import Compiled, Buffer, BufferOptions, CompiledASTRunner, BufferXfer, MultiDeviceJITGraph, update_stats
 from tinygrad.shape.symbolic import Variable
 from tinygrad.runtime.ops_hsa import HSADevice, PROFILE, Profiler
-from tinygrad.features.jit import JitItem, get_input_replace, get_jit_stats, \
+from tinygrad.engine.jit import JitItem, get_input_replace, get_jit_stats, \
                                   get_jc_idxs_with_updatable_launch_dims, get_jc_idxs_with_updatable_var_vals
 import tinygrad.runtime.autogen.hsa as hsa
 from tinygrad.runtime.driver.hsa import check, AQLQueue, AQL_PACKET_SIZE, EMPTY_SIGNAL
diff --git a/tinygrad/runtime/graph/metal.py b/tinygrad/runtime/graph/metal.py
index de54924ff3..49ab94c407 100644
--- a/tinygrad/runtime/graph/metal.py
+++ b/tinygrad/runtime/graph/metal.py
@@ -3,7 +3,7 @@ import Metal
 from tinygrad.dtype import dtypes
 from tinygrad.helpers import dedup, unwrap2, GraphException
 from tinygrad.device import Buffer, CompiledASTRunner, update_stats
-from tinygrad.features.jit import JitItem, get_input_replace, get_jit_stats, get_jc_idxs_with_updatable_launch_dims
+from tinygrad.engine.jit import JitItem, get_input_replace, get_jit_stats, get_jc_idxs_with_updatable_launch_dims
 from tinygrad.shape.symbolic import Variable
 from tinygrad.runtime.ops_metal import MetalDevice, wait_check
 
diff --git a/tinygrad/tensor.py b/tinygrad/tensor.py
index 79c1644ce6..31959c68c2 100644
--- a/tinygrad/tensor.py
+++ b/tinygrad/tensor.py
@@ -14,7 +14,7 @@ from tinygrad.features.multi import MultiLazyBuffer
 from tinygrad.ops import LoadOps
 from tinygrad.device import Buffer, Device
 from tinygrad.shape.symbolic import sint
-from tinygrad.realize import run_schedule, create_schedule
+from tinygrad.engine.realize import run_schedule, create_schedule
 
 # **** start with two base classes, Tensor and Function ****