At Onified.ai, I’m contributing to the AI architecture for a multi-tenant SaaS platform. I built its knowledge base ingestion pipeline, using structure-aware chunking and pgvector to power RAG retrieval.
I also implemented a provider-agnostic embedding layer through LangChain, with per-tenant configuration for Google Vertex, OpenAI, and Azure OpenAI. For agent configuration changes, I implemented transactional-outbox event publishing over Kafka to keep Redis invalidation atomic with database commits.
On LegalAgent, I built an agentic contract review pipeline that analyzes clause relationships and detects risks, with explanations grounded in RAG. I also developed VeriTas, a real-time fake news classification application using a fine-tuned RoBERTa model.
