Yash Pabari

About Yash Pabari

I specialize in taking AI from experimentation to production in high-stakes environments.As an AI/ML Engineer at Nomura, I work at the intersection of LLMs, NLP and financial services — engineering pipelines that process daily conversation records for AI governance, and building investigation tools using vector embeddings and semantic search.I own the full pipeline — data ingestion, model evaluation, concurrency optimisation, backend APIs, cloud deployment and monitoring. End to end.Core stack: Python, LangChain, Hugging Face, FastAPI, Spring Boot, Docker, Kubernetes and AWS.I\'m driven by solving real, high-stakes problems with intelligent systems — and making sure they\'re fast, scalable and production-ready.

Experience

1. Analyst(CDO)

   Nomura

   Jul 2024 — Present · Mumbai, IN

   Working as an AI/ML Engineer within the technology division at Nomura, focused on designing and deploying intelligent systems, LLM-powered pipelines and backend automation in a financial services environment.• Engineered an AI Governance platform (2LOD) ingesting daily user-LLM conversation records from AWS S3, evaluating PII, toxicity and bias using DeBERTa-v3, RoBERTa Toxicity Classifier and Social Media Fairness Classifier.• Implemented multithreading with producer-consumer architecture and semaphore-based concurrency in Python to parallelise evaluation across 100-record batches within a 24-hour SLA, reducing compute overhead through deduplication.• Trained a custom intent classification model using Scikit-learn for per-intent metric computation, enabling granular AI governance reporting.• Orchestrated data flow into Snowflake and PowerBI dashboards for reporting; used MSSQL for metadata storage and full audit trail of evaluation runs.• Built a POC legal hold investigation system — processed Outlook PST emails, constructed keyword indexes, generated embeddings in a vector database and ran multi-algorithm semantic search to surface evidence of non-compliant activity. Approved for full production development.• Developed a Spring Boot + MongoDB backend for trade data exception querying, restoration and management.• Built a Java utility for automated HCP file deletion using multi-filter logic (date, file type, size).

Education

• Bhartiya Vidya Bhavans Sardar Patel Institute of Technology Munshi Nagar Andheri Mumbai

  Bachelor of Engineering - BE, Information Technology