Bhavesh Patil Email & Phone Number

Bengaluru, Ka, In

Bengaluru, Karnataka, India

Building decision-making AI copilot

Bengaluru, Karnataka, India

• Worked on developing smart CRM platform for better sales insights and building strong client relations• Implemented Retrieval Augmented Generation(RAG) with open-source LLMs for event recommendation, persona identification.• Built multi-level summarization and topic extraction pipelines with LLMs on multi-channel client interactions data • Prototyped in-house automated tools based on GenAI for HR & compliance, domain research, employee onboarding, data analytics and mental health

Mumbai, Maharashtra, India

• Developed a non-parellel end-to-end singing voice conversion algorithm utilizing PPGs signals, acoustic reference encoders for high-fidelity voice conversion• Experimented with different acoustic reference features like mel spectrograms, HuBERT embeddings, and PPG-Mid to capture singer-independent musical information and enable cross-domain capabilities

Mumbai, Maharashtra, India

• Devised and optimized an end-to-end pipeline for face recognition, identification and person tracking• Obtained accuracy of 99.1% on face recognition and deployed the API using Flask and AWS Lambda• Built shortest path optimization algorithm for cost minimization for supply chain management

Mumbai, Maharashtra, India

• Developed an absurdity detection and filtration algorithm using NLP achieving 98.4% accuracy• Designed APIs using Flask and deployed with the use of Docker; performed API testing with Postman• Implemented context and keyword extraction with Latent Dirichlet Allocation (LDA)

Mumbai, Maharashtra, India

• Worked on building strategies and designing the technical product for agricultural process automation • Developed end-to-end pipeline for weed detection in farms and improvised its performance over time • Guided and mentored 3 interns over 4 months for product development related various technical tasks

Mumbai, Maharashtra, India

• Worked on time-series modeling for energy load prediction and optimization in smart buildings • Experimented with (S)ARIMA, Random Forest, Deep Belief Network and Neural Networks for prediction• Achieved R-squared error of 0.99 outperforming the previous model with CNN-LSTM architecture

Mumbai, Maharashtra, India

• Spearheaded a team of 4 conveners, responsible for overall design and maintenance of a web interface• Collaborated with editorial board, design team and data team for smooth transfer and ideation of content • Strategized the traffic on website using google analytics and observed 80% y-o-y increase in readership