← Back to work
TMA Solutions
RFP Assistant
Drafts sales proposals (RFP responses) and answers client questions straight from their own PDF, Word and Excel files. 79% of answers fully correct.
- Role
- AI Engineer
- Context
- TMA Solutions
- Period
- Jul 2024 – Dec 2024
Overview
AI-powered proposal generation and sales chatbot built on RAG over unstructured client documents.
Key achievements
Achieved 79% fully correct, 11% partially correct, and 10% incorrect answers on a reviewed set of 200 question-answer pairs.
Responsibilities
- Implemented a full RAG pipeline (data preprocessing, data store, retrieval, and generation) using Python and LangChain Community.
- Used Google Cloud Platform services for the vector database and MongoDB for full-text data.
- Implemented document processing for Excel, Word, and PDF files using Document AI combined with rule-based document style logic.
- Integrated Vertex AI and Gemini to build the Q&A chatbot.
- Built a microservices system isolating the main application backend from AI services, using Python Tornado and FastAPI.
- Deployed the application on Google Cloud Platform.
Tech stack
Languages & Frameworks
PythonFastAPITornadoReact
LLMs & AI
GCP services (Vertex AIGeminiDocument AI)