Junior AI Engineer Intern
Mako IT Lab
Dec 2025 – Mar 2026 · 4 months
Dual-Agent System
Hybrid RAG Pipeline
Full Observability
- Architected a dual-agent Sales Intelligence system using LangGraph — a deterministic Sales Agent pipeline (document extraction → query generation → retrieval → pitch synthesis) paired with a context-aware RAG Chatbot with conversation-aware query rewriting.
- Built a hybrid retrieval pipeline on Qdrant Cloud combining dense embeddings, sparse vectors, and a cross-encoder reranker with parent-child chunking to retrieve contextually relevant document sections at high precision.
- Integrated LangExtract for schema-based structured extraction from unstructured company profiles; generated multi-angle business queries per prospect to maximize recall across industry and strategic dimensions.
- Deployed a Streamlit frontend with real-time LLM token streaming, session persistence, and tool integrations including Firecrawl web scraper and BrightData Twitter scraper.
- Instrumented the full pipeline with OpenTelemetry tracing and custom latency metrics, enabling end-to-end observability across retrieval, reranking, and generation stages.