Varadh Kaushik Email & Phone Number

United States

Somerset, New Jersey, United States

In my current role, I am working on deploying and optimizing AI models in a production environment. My responsibilities include:- Experimenting with advanced prompt engineering techniques to enhance model accuracy and reduce error rates.- Fine-tuning input data structures (large JSON files) to improve the model's comprehension and overall performance, ensuring seamless interaction between data and AI systems.- Continuously refining model outputs by integrating error correction strategies and data manipulation approaches for optimized results in real-world applications.

Somerset, New Jersey, United States

▪ Comprehensive Model: Conducted an intensive study on leading Large Language Models, including ChatGPT, Claude, Bard, T5, FlanT5, Falcon, Llama, and Llama-v2, gaining in-depth insights into both open-source and proprietary AI technologies.▪ Cost-Effective Model: Streamlined operations and cut costs by $50K annually by transitioning from external APIs to custom in-house models, demonstrating fiscal responsibility and resource optimization.▪ Model Performance: Improved model performance by a notable 20% compared to the OpenAI API.▪ Dataset Development: Created fine-tuning datasets from internal company data, enhancing the applicability and relevance of ML models.▪ Optimized Model: Reduced model hallucinations by an impressive 70% through strategic fine-tuning, enhancing the reliability of AI outputs.▪ Quantization: Implemented strategic model quantization techniques, optimizing resource usage under computational constraints and contributing to efficient AI model deployment.▪ Autonomous AI: Pioneered the company's transition to a more autonomous AI infrastructure, playing a key role in shaping the future of AI operations within the organization.

Researched, prototyped and modified anomaly detection libraries to perform uni-variate and multi-variate anomaly detection on time series data.Some libraries which were researched include:- Facebook Prophet- LinkedIn Greykite- PyOD

Worked on a patented Augmentative and Alternative Communication [AAC] device which enables patients suffering from ALS, who have lost muscle control, to communicate via eye movement.My contributions included:- Creating and training a computer vision model to detect the current state of the eye- Work on the Android application to establish communication between the Raspberry Pi on the device and an Android device- Python scripting to perform operations on huge image datasets

Master Of Science - Ms, Artificial Intelligence, 3.93/4.00

Bachelor Of Technology - Btech, Computer Science