This portfolio showcases the engineering principles required to build reliable, high-stakes intelligent systems. I construct environments where AI moves past mere suggestion; it executes verified logic. Expect to see deeply architected solutions—from multi-agent fault tolerance to hybrid search that combines the precision of BM25 with the nuance of kNN.
Review these projects to see how deterministic engineering anchors powerful intelligence.
[Node 01] LinkedIn SSI Booster A truth-gated, continual learning automation agent. Driven by hybrid retrieval that leverages BM25 for deterministic keyword scoring, NetworkX for structural graph mapping, and spaCy for rigorous NLP semantic validation to strictly ground outputs. 🌟 Star linkedin_ssi_booster on GitHub 🌟
[Node 02] Regulatory Intelligence Assistant (RIA) Built for the G7 GovAI Grand Challenge, this multi-tiered search architecture masters deep indexing. It combines Elasticsearch, Neo4j graph traversal, and vector search to navigate massive federal legal datasets with absolute precision. 🌟 Star regulatory-intelligence-assistant on GitHub 🌟
[Node 03] Answer42 A sophisticated, 9-agent orchestration pipeline designed for academic research analysis. Features automated fallback logic and circuit breaker protection via Spring Batch—seamlessly shifting from cloud APIs to local Ollama models without dropping the execution thread. 🌟 Star answer42 on GitHub 🌟
Found in: linkedin_ssi_booster | General RAG Concepts Moving beyond simple prompt wrapping to combine deterministic fact-checking with advanced semantic retrieval.
- Hybrid Retrieval: Blending deterministic keyword ranking (BM25) with structured relationship mapping (NetworkX).
- Truth Gate: A multi-stage validation layer analyzing candidate text via BM25 evidence scoring and spaCy semantic similarity checks. This minimizes hallucination by grounding LLM outputs strictly in private, persona-defined knowledge.
- Continual Learning: The system actively tracks published output to refine source weighting and source reliability priors over time.
Found in: Answer42 | SIGNAL-ZERO | Grizzly News Building robust, resilient systems that require specialized roles to execute complex tasks.
- Agentic Pipelines: Architecting complex workflows (e.g., the 9-agent academic analysis pipeline in Answer42, or 5-agent detection systems) where specialized nodes communicate and refine outputs sequentially.
- Resilience & Local Routing: Implementing robust fallback mechanisms and Spring-managed circuit breakers to guarantee uptime. In strictly local environments, I utilize intelligent NLP routing to efficiently manage hardware constraints—ensuring traffic is directed seamlessly without overloading the GPU with multiple Gemma instances simultaneously.
- Protocols: Deep proficiency with modern inter-service communication (MCP/FastMCP).
Found in: Regulatory Intelligence Assistant | CIHR Portal Mastering structured, large-scale information retrieval across heterogeneous data sources.
- The Stack: Deploying sophisticated, multi-layered combinations: Elasticsearch (BM25), Neo4j (Graph and relationship traversal), and vector search (kNN).
- Search Logic: Implementing multi-tier search fallbacks. If one index fails or lacks context, the query degrades gracefully to the next most suitable mechanism, ensuring high relevance and sub-500ms latency.
- Domain Focus: Conquering the complexity of querying hundreds of thousands of federal legal documents.
Found in: TPG/USPS JMS Platform | Shared Services Canada
- High-Throughput Core: Extensive architectural experience building, maintaining, and scaling core Java enterprise systems and high-volume event-driven platforms.
Whether you're here for the code, the algorithmic soundscapes, or just to connect across the grid:
- 👤 LinkedIn: Shawn Jackson-Dyck 🚀
- 🎶 Suno: Rei Toei (@samjd42) 🚀
▶️ YouTube: @samjd42 🚀- 📸 Instagram: @samjd.42 🚀
Building truth-gated systems and producing tracks takes heavy compute cycles. If you want to support the late-night builds: