Building useful systems from idea to deployment.

ProblemA live AI system had 800 ms latency.

ActionFound the bottleneck and rewrote that part; no specific cause is claimed.

ResultLatency fell to under 200 ms, a 75% reduction.

ProblemCPU-only hardware was too slow for real-time work.

ActionDeployed FP16-quantized CNNs with TensorFlow Lite on NVIDIA Jetson Orin Nano using CUDA.

ResultMeasured 4×+ GPU speedup over a Raspberry Pi CPU baseline.

ProblemFarmers needed crop and soil diagnosis in the field.

ActionBuilt a Python and Django REST system with IoT sensors and simple farmer dashboards.

Result96%+ disease detection, used at 3+ farming sites.

ProblemData work was manual and repeated.

ActionAutomated 5+ recurring data pipelines with Python and cleaned and validated large datasets.

ResultLess manual work; received a Letter of Recommendation.