About Me

Dushyanta is a Senior Applied Scientist at Amazon, where he is the technical lead for Amazon Pricing’s Foundational modeling team, re-imagining price estimation for a billion-scale product catalog. His research spans information retrieval, LLM-based probabilistic regression, efficient LLM inference, and agentic systems for automated auditing. He previously led the science strategy for Amazon Books’ Inauthentic Book Duplicates and Content Moderation programs, and joined Amazon as a Machine Learning Engineer on Alexa Brain and Amazon Comprehend.

Dushyanta holds an MS in Computer Science from The Ohio State University, where he worked with Prof. Huan Sun on Weakly Supervised methods for Neural Relation Extraction , and was previously a Research Assistant at the Ubiquitous Knowledge Processing Lab under Prof. Dr. Iryna Gurevych .

Dushyanta’s Resume

• (September, 2026) One paper was accepted to the EMNLP 2026 Main Conference, and another to the AKBC 2026 Workshop

• (April, 2026) Two papers were accepted to ACL 2026: Text-to-Distribution Prediction with Quantile Tokens and Neighbor Context and Breaking the Autoregressive Chain: Hyper-Parallel Decoding for Efficient LLM-Based Attribute Value Extraction (Findings)

• (June, 2025) My paper on LLM-based distributional regression was accepted to Findings of ACL 2025

• Senior Applied Scientist, Amazon Pricing (Sep 2024 – Present)

  Technical lead for a cross-functional team, re-imagining Amazon’s product price estimation strategy.

  • Developed a retrieval-augmented system for price distribution estimation using similar-product pricing signals at billion-scale across marketplaces and locales.
  • Launched an LLM-based price quality assessment system for channel-agnostic pricing.
  • Developed a hyper-parallel LLM attribute extraction pipeline achieving 90%+ inference cost reduction.
  • Guided development of a centralized MLOps hosting and deployment platform.
  • Conceptualized and implemented Evaluation Driven Science, a framework for grounding roadmap prioritization in systematic model evaluation.

• Senior Applied Scientist (Apr 2024 – Aug 2024), Applied Scientist II (Oct 2020 – Mar 2024), Applied Scientist I (Oct 2019 – Sep 2020), Amazon Books

  • Started and led a cross-functional team to build Content Moderation solutions for Amazon Books, including:
    • Hate speech and explicit content detection: Developed multi-task, multi-instance learning (MIL) approaches to detect policy-violating content in long-context book interiors.
    • Copyright infringement detection: Designed and implemented an end-to-end, multi-stage text ranking system, including a transformer-based neural ranking model for book-pair scoring, a scalable hybrid retrieval engine combining lexical and vector search, and a near-real-time inference service processing 500M book pairs per day.

---

• Research Assistant, The Ohio State University

  Worked under the supervision of Prof. Huan Sun on developing techniques for effectively utilizing clean and noisy data in natural language processing.

• ML/AI Frameworks: PyTorch, PyTorch Lightning, HuggingFace Transformers, numpy, scikit-learn
• LLM & NLP: LLM fine-tuning, SFT, RAG, information retrieval, probabilistic regression, efficient inference
• Large-Scale ML Systems: PySpark, Airflow, AWS SageMaker (distributed training and inference), AWS Trainium