Anish Kumar Email & Phone Number

Bengaluru, Karnataka, India

At Synamedia, I gained hands-on experience with CDN 🌐 and streaming technologies, focusing on optimizing global video delivery. I started by analyzing the SNOW dataset, extracting valuable insights, and applying transfer learning to LLMs for more efficient query processing.I developed a RAG-based query bot 🤖 using FASSI indexing, significantly improving user experience by delivering accurate, real-time responses. Additionally, I designed and implemented a chatbot flow, seamlessly integrating it with a scalable REST API to process user inputs and enhance service interactions.In addition to development, I worked on containerizing services 🧑‍💻 and automating deployment through CI/CD pipelines, enabling faster and more reliable updates. Throughout, I collaborated with cross-functional teams, ensuring high-quality, scalable solutions that drove innovation and efficiency in the media domain.

Bengaluru, Karnataka, India

At On-Demand Experts, I played a key role in developing a customer-facing web app 💻. I began by gathering product requirements and creating a design document 📑, incorporating customer feedback to define features and technology choices.I designed the UI in Figma 🎨 for maximum user engagement, then developed the prototype using Next.js, AWS Amplify, and Facebook API, with a focus on SEO. After deployment, I used Google Analytics to track user behavior 📊, iterating on feedback to improve the app. The final product received positive investor feedback.

Bengaluru, Karnataka, India

At Samsung, my team and I started with limited experience in AI/ML 🤖, so we spent time learning the technologies before progressing toward training a vision model. A significant portion of our effort went into data collection, cleaning, augmentation, and labeling to ensure a solid dataset for model development.After extensive trials, we chose YOLO Tiny v4 for its accuracy, inference speed, and compatibility with our needs. We optimized the model, fine-tuned parameters for maximum accuracy, and proceeded with quantization to deploy it on an Android system 📱.The next challenge was creating a stable Android application with limited device resources. We developed a real-time object counting algorithm using cosine similarity to track and distinguish objects in video streams 🎥. This required close collaboration with the Samsung team, and under their guidance, we successfully documented the novel algorithm we developed for future use.