[ad_1]
Earlier than finishing PGP-AIML in Nice Studying, I used to be purely in an Infrastructure background on backup and storage. At present, I’m not carrying any non-technical expertise. My whole expertise could be 16+ and 14+, which was totally within the backup and storage area. I used to be a backup and storage administrator earlier than becoming a member of this program within the 2019 batch.
The issue which I confronted at my office, the place I utilized Synthetic Intelligence or Machine Studying at Work – “Deep studying initiatives had been occupying extra space and ram consumption whereas working the undertaking. There was an Error prevalence on the information annotation half, which impacted the mannequin efficiency.”
Whereas working the undertaking on the Pod man container, the dimensions of the Keras was too large to suit. Therefore, there was a necessity to cut back the dimensions of the .H5 file to suit on the Containers. Many handbook errors have occurred whereas performing the information annotation on the textual content and pictures. This was impacting on-time supply.
Instruments & methods that I utilized to unravel the issue:
Keras Quantization and Distillation had decreased the trainable parameters and the dimensions of the .H5 information. This helped us to construct the pod principal picture to ship the undertaking. Automated validation on the Annotated samples to cut back the handbook errors. This decreased the bias on the mannequin improvement and improved the efficiency. In flip, it helped out the undertaking deliverable on time.
The Insights & Options beneficial by way of these AI/ML idea purposes had been – In-depth evaluation and validation.
The Influence my suggestions generated on the group –
Enchancment within the undertaking supply on time
This train benefitted me in upskilling myself on daily foundation. Nice Studying’s Mentored Studying Classes, their Program Help, and the Video Content material, together with the Studying Supplies, helped me all through this journey within the AIML Program.
[ad_2]
