Pouya Javadi

Innovative Software Engineer with expertise in React development, GraphQL APIs, and Python programming. Proven track record in driving impactful projects and optimizing backend infrastructure to ensure scalability and reliability.

Vector Search RAG System for Unstructured Data
Personal Project github.com/pjavadi84/new-ai-project.git
• Designed and implemented a Retrieval-Augmented Generation (RAG) pipeline using Python to enable Large Language Models (LLMs) to query and synthesize answers from unstructured documents (e.g., PDFs).
• Developed vector search indexing for document content, leveraging high-dimensional embeddings and a
vector database (e.g., Chroma/FAISS) to ensure fast, accurate retrieval and context grounding.
• Established end-to-end data processing workflows for efficient chunking, embedding, and storage of documents, significantly expanding the LLM’s effective knowledge base.

Symbol Machines: Cognitive AI Architecture
Collaborative Research Project github.com/symbolmachines/sdk.git
• Collaborated on the design and implementation of a ”Digital Hippocampus” architecture to provide stateful, persistent memory for conversational AI, addressing the limitations of temporary context windows.
• Developed the core retrieval logic for a 5-Channel Hybrid Retrieval system (incorporating Semantic, Emo-
tional, and Temporal data) to enhance memory recall relevance and cognitive continuity.
• Engineered a Dual-LLM Topology to separate user interaction (fast response) from background memory
consolidation (asynchronous graph updates in Neo4j), significantly improving chat latency.
• Integrated Valence-Arousal-Dominance (VAD) vectors for emotional indexing of memories, allowing the system to retrieve information based on affective resonance, similar to human memory.