Akshay Kumar Arasavalli

1. Fine-Tune Mistral-7B using LoRa on custom dataset for Taxonomy generation for Document 2. Spearheaded prompt engineering and analysis of multiple LLM models, including Falcon, Dolly, GPT, and MPT, to identify the most effective model for the given project requirements. 3. Collaborated with cross-functional teams to establish project objectives, gather requirements, and define timelines for real-time auction bidding optimization in first-price auctions. 4. Designed and developed scalable and efficient machine learning (ML) models to enhance bidding strategies and improve auction performance. 5. Utilized reinforcement learning (RL) algorithms and simulation techniques to train intelligent bidding agents capable of adaptive decision-making in dynamic first-price auction environments. 6. Stayed updated with the latest advancements in ML, AI, and RL research and technologies, incorporating cutting-edge insights and best practices into the auction optimization process. 7. Tech Stack: Python, PyTorch, hugging face, Django, Sagemaker, Snowflake

1. Designing AI systems and developing them from scratch, leading a team of engineers and data scientists. 2. Scaling the architecture by moving from a monolith to microservices, building low-latency systems at scale. 3. Leading efforts to improve overall infrastructure, from CI/CD to cloud cost optimization. 4. Mentoring team members, devising roadmaps, charters, and visions for the team 5. Tech Stack: Python, Fasttext, Pandas, Django, RabbitMQ, AWS Glue, AWS Athena, EKS

Targeting AI: 1. To generate relevant audience cohorts and expand the audience based on learning using state-of-the-art Language Modeling and Machine Learning techniques. 2. Once the Targeting AI is deployed and activated, it monitors performance and creates sharper and stronger targeting parameters that can help achieve campaign goals faster. The beauty of this AI system lies in its ability to ingest and analyze unbelievable amounts of data within seconds and provide targeting recommendations that are simply impossible for us to arrive at on our own, at that speed. 3. Supported Platform: Facebook Ads, Google Ads 4. Tech-Stack: - Python, Fasttext, Pandas, Django, Rabbitmq, AWS Glue, AWS Athena, EKS

1. Worked towards the vision of Jotter.ai, designing and developing deep-learning image processing solutions. 2. Transformed images from one pose to another using CNNs and image processing techniques. 3. Implemented deep auto-encoders for high-res (4k) image segmentation, worked with various object detection and segmentation architectures. 4. Created a custom segmentation architecture based on ResNet-50 (encoder), custom decoder, and UNet style lateral connections. 5. Developed an AI segmentation network to extract apparel from raw DSLR images. 6. Trained a model with a huge amount of data (~1TB) on a GPU cluster. 7. Tech Stack: Python, OpenCV, TensorFlow, Docker, Tensor RT