I am Harinath Reddy Cingapuram, a research-focused AI/ML engineer and data scientist with experience across Fortune 500 companies and high-impact academic projects in generative AI and large language models. My work spans transfer learning, LLM alignment, and retrieval-augmented generation, with multiple peer-reviewed publications and ongoing conference submissions in applied machine learning and AI systems. I am particularly interested in building scalable, cloud-native ML pipelines and collaborating on research that bridges cutting-edge theory with real-world deployment on platforms such as AWS, Azure, and OpenShift.
My research focuses on advancing large language models through alignment techniques transfer learning and retrieval-augmented generation. I am particularly interested in bridging theoretical advances in generative AI with practical deployment challenges in cloud-native environments. My work spans supervised fine-tuning of LLMs diffusion models for image generation and building scalable ML pipelines on distributed systems. I am also exploring federated learning approaches and the intersection of AI systems with real-world applications in healthcare finance and enterprise solutions.
Large language models often plan solutions correctly yet stumble within individual reasoning steps, especially when guided by coarse, problem-level demonstrations. This paper introduces a probe- and-guide in-context learning framework that aligns guidance to the step granularity: the model first issues a brief probe for the next step, then retrieves and conditions on closely matched example steps from a cu- rated repository to execute the step with higher fidelity. The approach reduces irrelevant-example noise, improves single-step correctness with- out additional training, and slots into standard inference pipelines and search-based decoders, enhancing both candidate generation and verifi- cation. Evaluations across diverse mathematical reasoning settings show consistent gains over zero-shot and few-shot baselines, and the method composes naturally with tree-search strategies to further improve solu- tion quality while controlling token cost. The design is model-agnostic, training-free, and centers on LLM inference workflows, making it practi- cal for deployments that demand reliable, fine-grained reasoning.
Looking for researchers interested in advancing large language model alignment techniques, reinforcement learning from human feedback (RLHF…
Looking for researchers interested in advancing large language model alignment techniques, reinforcement learning from human feedback (RLHF…