Udit Mittal

Built a RAG-based message classification system using FastAPI, Weaviate, and OpenAI with custom embeddings and hybrid search strategies, processing 2M+ messages for intent classification and routing, reducing agent workload by 40%

Engineered an automated form digitization system using custom-trained Azure OCR models and validation algorithms,increasing form processing throughput by 350% while maintaining 90% accuracy. Reduced manual data entry time by 75% through robust error handling, automated validation, and comprehensive monitoring using Langfuse for OCR outputs and processing steps

Developed a RAG app with Next.js and React.js, using Weaviate for vector search to enhance LLMs like OpenAI with external data. Integrated with Vercel AI SDK and integrated real-time monitoring and logging via Langfuse

Architected a real-time language detection pipeline using FastAPI and Exotel, integrating Sarvam AI’s API withIVR system to reduce language selection time by 80%.

Revamped existing ad-hoc data serialization with Django Rest Framework, decoupling it from data models for improved code maintainability, robustness, and reduced integration effort across 10+ frontend components

Improved code quality by writing a robust test suite using the Python library, PyTest to ensure stability andmarketability of critical modules

Led a successful migration of CI/CD pipelines from Bitbucket Pipelines to GitHub Actions, optimizing deployment efficiency. Additionally, established a streamlined CI pipeline for frontend integration testing, resulting in a 20% reduction in integration time.