We are looking for a talented and experienced Data Scientist to join our team at Yum India Global Services Private Limited in the BMU YC India department. In this role, you will play a crucial part in shaping the future of our AI-powered systems, particularly in the realm of conversational AI. Your expertise will be instrumental in refining prompt instructions, designing conversation flows, and building robust error-handling routines, all while maintaining the unique brand voice of Taco Bell. Additionally, you will contribute to model tuning and evaluation, collaborating closely with our MLEs, QAs, and PMs to enhance agent usability, reliability, and task success rates.
* Author and refine prompt instructions, chaining logic, and fallback strategies to ensure seamless conversation flows.
* Design and test multi-turn conversation flows, aligning them with the Taco Bell brand voice and ensuring a consistent user experience.
* Build and maintain system personas and error-handling routines to handle unexpected user inputs and maintain a smooth conversation flow.
* Fine-tune LLMs, ASR models, and embedding systems under supervision, optimizing their performance and accuracy.
* Assist in running experiments using various methods such as LoRA, distillation, or pruning to enhance model efficiency and effectiveness.
* Contribute to agent evaluation metrics, regression tracking, and A/B tests, ensuring the continuous improvement of our conversational AI systems.
* Work closely with MLEs on model integration and performance tuning, ensuring a seamless collaboration between different teams and expertise.
* Partner with QA and PMs to improve agent usability, reliability, and task success rates, ensuring a positive user experience and efficient task completion.
* Help manage RAG components, context retrieval chains, and structured data inputs, ensuring the system can access and utilize relevant information effectively.
* 4-8 years of experience in AI Engineering, Data Science, or ML-related roles, with a strong background in conversational AI and NLP.
* Proficiency in Python, SQL, and AI frameworks such as LangChain, HuggingFace, and OpenAI APIs, ensuring a solid foundation for developing AI solutions.
* Hands-on experience in LLM/NLP fine-tuning, including SFT, LoRA, QLoRA, and PEFT frameworks, to optimize model performance and accuracy.
* Strong experience with RAG development, ensuring the system can effectively retrieve and utilize relevant information for conversation.
* AWS proficiency, including S3, Lambda, API Gateway, ECS/EKS, and SageMaker, to deploy and manage AI models and applications.
* Ability to convert models into production-ready applications, including API creation, microservices, Dockers, CI/CD pipelines, and Kubernetes.
* Experience in building data/ML pipelines for transcripts, call logs, and conversation data, ensuring efficient data processing and analysis.
* Comfortable working with US engineering teams, facilitating effective cross-timezone collaboration and ensuring smooth project execution.
* Exposure to ASR/TTS outputs and voice-to-text workflows, providing an understanding of voice-based interactions and their unique challenges.
* Understanding of Conversational AI KPIs such as containment, handoff/fallback, and AHT impact, ensuring the system meets key performance indicators.