Ashish Kumar Sinha Email & Phone Number

Bengaluru, Karnataka, India

Banglore

- Developing advanced GenAl solutions using Open Sourced LLMs like Mistral, LLAMA, etc. to build innovative solutions. My experience extends to the application of advanced techniques like Prompt Engineering and Retrieval Augmented Generation (RAG) to fully capitalize on the potential of LLMs. I have further employed Langchain, VectorDB, embeddings, and hybrid & semantic search (Information Retrieval) techniques to enhance solution effectiveness. The main technologies I am using are Python, Langchain, FAISS, and ChromaDB.- Developed a user segmentation model based on personal system usage for 20M users.- Employed various models, including LSTMs and self-attention mechanisms.- Made robust and explainable models by implementing Evidential Deep Learning and SHAP.- Collaborated with other teams, such as AB-testing and RCTs, to optimize product recommendationsand increase revenue.- Streamlined data processing and storage by using AWS services such as EMR, EC2, and S3

- Developed large scale user segmentation model utilizing innovative deeplearning techniques to enhance HP DaaS solutions.- Implemented hierarchical RNN models to improve time series forecasting accuracy.- Managed and optimized the handling of sparse usage data through multidimensional tensor implementation.

- The project aimed to build and deploy a novel deep learning model that predicts time to a certain critical event and was undertaken as a semester-long industry project.- Worked with HP proprietary data and fit time-to-event prediction model on it.- Experimented with variations of Deep Recurrent Survival Analysis model, build various sequential modelsand test their predictive capabilities.- Worked with a team of professionals to design and implement training and inference pipelines for the model that would ultimately be pushed into production.

- Worked on time-to-event sequential modelling to predict print failure for large scale HP printers- Gained familiarity with Keras API to develop Deep Learning Models in Python.- Wrote custom loss functions for DRSA model and Batch Generators in Keras for efficient memory handling for large datasets.- Experimented with various RNN models including Stateful RNN for the temporal dataset.- Experimented with Cyclic Learning Rate and One Cycle policy to speed up the convergence of the model.

Bangalore, India

The project aims to find the patterns from markers on a deformed surface using image pre-processing and Artificial Neural Networks.OpenCV was used to for image processing and the Machine Learning Model was developed using Keras framework. Convolutional Neural Network was used to classify the markers on the deformed surfaces.

Master Of Science (M.Sc.), Data Science, 8.69

Bachelor Of Science (B.Sc.) (Hons.), Statistics(Hons.),Mathematics & Computer Science, 8.78

Higher Secondary Examination

Secondary School Certificate

Junior Diploma, Stringed Instruments(Violin), A