Machine Learning Engineer

Created and assisted with the development of 5 chatGPT-based custom service chatbots for Business Intelligence(BI) and other commercial purposes like banking and deployment of them into mobile apps and web servers.
Collaborated with product, engineering, and customer teams under Agile methodolgy to design, build, improve and fine-tune ML models(Large Language Models) tailored to specific enterprise use cases.
Developed and deployed a DL model for a financial services client, resulting in a 25% reduction in fraudulent transactions and saving the client $50k annually.
Constructed 3 different deep neural networks and generated, new, realistic faces using GAN(Generative Adversarial Network) on PyTorch environment.

Deployed 2 attention-based encoder-decoder models to check grammar error.
Trained 2 Chinese-English machine translation systems that outperformed Google Translation using transformers and some manually organizing models.
Spearheaded transition from firebase to AWS, saving company $27000+ monthly and increasing load speeds by an average of 42%.
Initiated a character-based transformer spelling correction model using Triton and ONNX; reduced inference latencies below 178ms at 35 req/s.
Implemented authentication/authorization mechanism using Auth0 and AWS Cogito.
Developed an NLP system that automatically classified 7.5K emails as spam or advertising mail using 3+ natural language processing methods.

Trained a CNN(Convolutional Neural Network) to analyze images of dogs and correctly identify their breeds and used transfer learning to increase performance by 5% and simplify the model.
Analyzed and interpreted data for a retail client to identify trends and patterns for 2 years.
Applied LSTM network to predict the pattern Assisted in creating a DL model that detects anomalies in the building process of a construction robot and machine reducing downtime by 83% & processed action measures to predict the state of machine reducing fix cost by 20%.
Conceived and discovered a machine learning algorithm that detects deviant behavior in robots using SIFT, HOG, and 20+ other computer vision methods.
Tracked the health of 15+ robots using React/Redux with NodeJS backend and Python scripts, which collected 1TB of data from its sensors.