Shastri Vaishampayan Mohapatra

• Architected CFG-driven synthetic data generation pipelines to scale training corpora for transformer-based NLU models, improving intent classification and NER generalisation across multilingual (Hindi, English), automotive voice assistant interactions.
• Led data-centric model optimisation cycles through fine-grained error analysis and metric-driven evaluation (F1, accuracy), resulting in measurable improvements in intent disambiguation, and entity extraction performance in production systems.
• Designed and deployed annotation quality assurance frameworks using regex and rule-based validation, reducing schema violations, label noise, and strengthening training data reliability for large-scale NLP pipelines.
• Conducted linguistically informed morphological analysis for Hindi, enabling the development of robust normalisation and subword tokenisation strategies aligned with transformer tokenisation schemes.
• Engineered tokenisation and normalisation grammars to improve handling of noisy, code-mixed, and morphologically rich inputs, leading to enhanced NLU robustness and downstream interpretability.
• Contributed to improving end-to-end speech-to-intent understanding by optimising text normalisation layers bridging the ASR output and downstream NLU components, resulting in better alignment between spoken input variability and structured intent parsing.
• Collaborated across ASR, NLU, and product systems to optimise the speech-to-intent pipeline, ensuring alignment with real-world conversational distributions, long-tail queries, and edge-case behaviours.

• Collaborated on LLM and Generative AI projects, focusing on dataset curation, annotation, and linguistic analysis for conversational systems.
• Performed LLM evaluation, prompt annotation, and response summarisation to improve generation quality and conversational coherence.
• Contributed to linguistic resources and data pipelines supporting conversational AI and language generation models.
• Conducted error analysis and system evaluation, driving improvements in model performance and reliability.
• Partnered on high-quality dataset creation and QA for GenAI systems, ensuring robust training data.
• Mentored junior linguists on annotation standards and LLM evaluation practices.
• Applied PII masking and text normalisation aligned with Responsible and Ethical AI principles.

• Developed a phonemiser for Odia, mapping native phonemes to IPA to support ASR, TTS accuracy, and speech processing pipelines.
• Contributed to conversational AI systems (voice bots and WhatsApp chatbots) by delivering high-quality multilingual data (English, Hindi, Odia) for NLU, ASR, and TTS in the BFSI domain.
• Designed conversation flows and mind maps, optimising dialogue structure and user experience across voice and chat interfaces.
• Applied conversation design principles to enhance customer experience in use cases such as debt collection, sales, and onboarding.
• Served as domain-specific POC for data quality, overseeing quality analysis, team coordination, and output consistency across annotation workflows.

• Performed translation, editing, and review (English–Oriya, Hindi–English), delivering client-facing, production-quality content across academic, healthcare, industry, application, and news domains.
• Conducted QC/QA and evaluation of translation and voice-over outputs, including MT/NMT systems, ensuring linguistic accuracy, domain adaptation, and consistency.
• Guided reviewers and enforced quality standards and best practices, ensuring high-quality, domain-aligned deliverables across diverse content types.

Development of a rule-based multi-lingual parser and machine translation system.

• Designed a rule-based multilingual parsing and machine translation system for translating Hindi into Oriya, English, Tamil, German, and Japanese.
• Developed core linguistic resources, including a transfer grammar module, semantic mapper, and multilingual concept lexicon, enabling structured cross-lingual generation.
• Conducted semantic analysis, annotation, and disambiguation to build a Universal Semantic Representation (USR), improving translation consistency and accuracy.

• Developed and curated high-quality synthetic Hindi datasets for Conversational AI (chatbots) in the BFSI domain, improving intent understanding and response relevance.
• Created gold-standard datasets for query disambiguation, enhancing the chatbot’s ability to accurately interpret complex, ambiguous user inputs.
• Designed hybrid (synthetic + gold) QA datasets for banking chatbots, enabling robust conversational performance and efficient query resolution.

• Worked on transcription and transliteration for Oriya, Hindi, and Sanskrit, supporting accurate text representation and basic disambiguation tasks.
• Performed quality evaluation of Indian English pronunciation, including phonetic analysis for consistency and correctness.