Senior Software Engineer at Stack AI

Led the development and operation of the multi-tenant Knowledge Base service with CRUD/version APIs in FastAPI, backed by PostgreSQL and Redis for state and queueing. Built a retrieval layer combining pgvector dense search with BM25 keyword matching, metadata filters, and rerank passes, reducing average LLM context tokens by 35% and increasing top-1 accuracy by 12 percentage points in internal evals. Engineered a resilient indexing pipeline parsing uploads from S3 (PDF, DOCX, CSV) with configurable chunkers and optional OCR; parallelized ingest flows cut p95 indexing time from 90s to 25s. Delivered an embeddings layer with provider routing (OpenAI, Azure, BERT, PaLM2) depending on tenant policy, achieving ~22% latency reduction with no recall loss. Integrated SharePoint, Google Drive, and Jira connectors, pushing ingested content through the same parsing/embedding pipeline. Introduced a Tool Invocation mechanism to defer retrieval until explicitly called, reducing KB queries per chat s

Senior Full Stack Engineer at Vectara

Led the development of Vectara's open-source create UI package for scaffolding React-based GenAI chatbot interfaces with integrated API authentication. Architected responsive UI components for the RAG console with real-time query processing, inline citations, and hallucination mitigation. Implemented frontend integrations for data upload/indexing workflows, supporting multi-format ingestion and low-latency development environments. Built the React-Chatbot toolkit and vectara-ui library (TypeScript/Next.js) with multilingual support and SSR optimization for Vercel, empowering 10,000+ developers. Engineered ingestion backend for multi-format semantic indexing (PDF, HTML, DOCX, audio, JSON) with real-time OCR and speech-to-text via Python and Kafka, achieving end-to-end latency under 1.2 seconds per file. Implemented cross-lingual query support using multilingual encoders and dynamic embedding routing; contributed to Hallucination Evaluation Module (FCS). Designed and deployed a query re-

Software Engineer at Aisera

Engineered an AI-driven HR chatbot (HR Copilot) to automate onboarding, leave requests, and policy queries. Designed conversational flows with contextual memory and multi-intent recognition. Built backend orchestration using Django and Kafka to automate HR ticketing and approvals; authored automation scripts in Python and JavaScript to streamline repetitive workflows. Integrated Workday, SAP, and ADP APIs to enable self-service HR actions; implemented OAuth-based authentication and fine-grained RBAC for enterprise security. Built real-time HR dashboards with React and D3.js and created Cypress E2E test suites covering 85% of critical workflows for release automation.