diff --git a/Dockerfile.jetson b/Dockerfile.jetson
index bd4fe57..57b4fd9 100644
--- a/Dockerfile.jetson
+++ b/Dockerfile.jetson
@@ -6,7 +6,6 @@ RUN apt-get update && apt-get install -y \
     python3 python3-pip python3-dev gcc \
     libgl1 libglib2.0-0 \
     python3-libnvinfer python3-libnvinfer-dev \
-    python3-pycuda \
     && rm -rf /var/lib/apt/lists/*
 
 
diff --git a/_docs/02_tasks/done/AZ-180_jetson_orin_nano_support.md b/_docs/02_tasks/done/AZ-180_jetson_orin_nano_support.md
index cf45f55..8d2a5a2 100644
--- a/_docs/02_tasks/done/AZ-180_jetson_orin_nano_support.md
+++ b/_docs/02_tasks/done/AZ-180_jetson_orin_nano_support.md
@@ -18,7 +18,7 @@ The detection service cannot run on NVIDIA Jetson Orin Nano for two reasons:
 ## Outcome
 
 - A `Dockerfile.jetson` that builds and runs on Jetson Orin Nano (aarch64, JetPack 6.x)
-- A `requirements-jetson.txt` that installs Python dependencies without pip-installing tensorrt or pycuda
+- A `requirements-jetson.txt` that installs Python dependencies without pip-installing tensorrt
 - A `docker-compose.jetson.yml` with NVIDIA Container Runtime configuration
 - `convert_from_source()` in `tensorrt_engine.pyx` extended to accept an optional INT8 calibration cache path — if the cache is present, INT8 is used; otherwise FP16 fallback
 - `init_ai()` in `inference.pyx` extended to try downloading the calibration cache from the Loader service before starting the conversion thread
@@ -28,7 +28,7 @@ The detection service cannot run on NVIDIA Jetson Orin Nano for two reasons:
 
 ### Included
 - `Dockerfile.jetson` using a JetPack 6.x L4T base image with pre-installed TensorRT and PyCUDA
-- `requirements-jetson.txt` derived from `requirements.txt`, excluding tensorrt and pycuda
+- `requirements-jetson.txt` derived from `requirements.txt`, excluding tensorrt and installing PyCUDA via pip where the JetPack apt package is unavailable
 - `docker-compose.jetson.yml` with `runtime: nvidia`
 - `tensorrt_engine.pyx`: extend `convert_from_source(bytes onnx_model, str calib_cache_path=None)` — set `INT8` flag and load cache when path is provided; fall back to FP16 when not
 - `inference.pyx`: extend `init_ai()` to attempt download of `azaion.int8_calib.cache` from Loader before spawning the conversion thread; pass the local path to `convert_from_source()`
@@ -101,7 +101,7 @@ Note: AC-2, AC-5, AC-6 require physical Jetson hardware and cannot run in standa
 
 ## Constraints
 
-- TensorRT and PyCUDA must NOT be pip-installed — provided by JetPack in the base image
+- TensorRT must NOT be pip-installed — provided by JetPack in the base image. PyCUDA may be pip-installed on `l4t-jetpack:r36.4.0` because `python3-pycuda` is unavailable in the apt repositories.
 - Base image must be a JetPack 6.x L4T image — not a generic CUDA image
 - Calibration cache download failure must be non-fatal — log a warning and fall back to FP16
 - INT8 conversion and FP16 conversion produce different engine files (different filenames) so cached engines are not confused
@@ -114,7 +114,7 @@ Note: AC-2, AC-5, AC-6 require physical Jetson hardware and cannot run in standa
 
 **Risk 2: PyCUDA availability in base image**
 - *Risk*: Some L4T images do not include pycuda
-- *Mitigation*: Fall back to `apt-get install python3-pycuda` or source build with `CUDA_ROOT` set
+- *Mitigation*: Fall back to pip source build with `CUDA_ROOT` set when no `python3-pycuda` apt package is available
 
 **Risk 3: INT8 accuracy degradation**
 - *Risk*: Without a well-representative calibration dataset, mAP may drop >1 point
diff --git a/_docs/04_deploy/containerization.md b/_docs/04_deploy/containerization.md
index 8373f5d..5bc592b 100644
--- a/_docs/04_deploy/containerization.md
+++ b/_docs/04_deploy/containerization.md
@@ -121,7 +121,7 @@ Already exists: `e2e/docker-compose.test.yml`. No changes needed — supports bo
 |--------|--------------|
 | Base image | `nvcr.io/nvidia/l4t-jetpack:r36.4.0` (JetPack 6.2.x-compatible, aarch64) |
 | TensorRT | Pre-installed via JetPack — `python3-libnvinfer` apt package (NOT pip) |
-| PyCUDA | Pre-installed via JetPack — `python3-pycuda` apt package (NOT pip) |
+| PyCUDA | Installed via pip in `requirements-jetson.txt` because `python3-pycuda` is not available in the `l4t-jetpack:r36.4.0` apt repositories |
 | Build stages | Single stage (Cython compile requires gcc) |
 | Non-root user | `adduser --disabled-password --gecos '' appuser` + `USER appuser` |
 | Exposed ports | 8080 |
@@ -129,7 +129,7 @@ Already exists: `e2e/docker-compose.test.yml`. No changes needed — supports bo
 | Runtime | Requires NVIDIA Container Runtime (`runtime: nvidia` in docker-compose) |
 
 **Jetson-specific behaviour**:
-- `requirements-jetson.txt` derives from `requirements.txt` — `tensorrt` and `pycuda` are excluded from pip; TensorRT and PyCUDA are installed from the JetPack/L4T apt packages in `Dockerfile.jetson`
+- `requirements-jetson.txt` derives from `requirements.txt` — `tensorrt` is excluded from pip and installed from the JetPack/L4T apt packages in `Dockerfile.jetson`; PyCUDA is installed via pip on this image line because the apt package is unavailable
 - Engine filename auto-encodes CC+SM (e.g. `azaion.cc_8.7_sm_16.engine` for Orin Nano), ensuring the Jetson engine is distinct from any x86-cached engine
 - INT8 is used when `azaion.int8_calib.cache` is available on the Loader service; precision suffix appended to engine filename (`*.int8.engine`); FP16 fallback when cache is absent
 - `docker-compose.jetson.yml` uses `runtime: nvidia` for the NVIDIA Container Runtime
diff --git a/requirements-jetson.txt b/requirements-jetson.txt
index fac7843..79fb685 100644
--- a/requirements-jetson.txt
+++ b/requirements-jetson.txt
@@ -12,3 +12,4 @@ requests==2.32.4
 loguru==0.7.3
 av==14.2.0
 xxhash==3.5.0
+pycuda