_{Switch language / 언어 전환}

Technical AI Product Manager & Service Architect > I don't just write product requirements; I build prototypes to prove them.
My career started in operations, business planning, and data analysis. Frustrated by the gap between ambiguous business needs and technical realities, I started building AI architectures and automation workflows myself.
Today, I design AI services, build Proof-of-Concepts (PoCs) to test feasibility, validate them with data, and make technical "Go/Drop" decisions based on real-world constraints.

"Memory is not a log. Memory is compacted meaning."

I am not a core AI researcher who optimizes low-level memory or derives mathematical formulas from scratch. My weapon is execution and architectural thinking. When faced with a business bottleneck, I combine existing models, APIs, and algorithms to test if an idea actually works in a constrained business environment. I build to find structural limits, and I use data to objectively decide whether to pivot, scale, or kill a project.

• Production First: I treat AI systems as operational systems, not isolated model demos.
• Embrace Constraints: I optimize for cost, stability, and integration friction, not just raw accuracy.
• Data-Driven Decisions: If a system doesn't prove its worth through metrics (e.g., speed, pass rate), I objectively document the failure and kill it.
• Learn from Limits: I turn failures, trade-offs, and dead ends into structured post-mortems and better system designs.

• 🧠 Agent Memory System MCP-based long-term memory architecture.
  Architecture: Strictly separated the State of Truth (MongoDB) from the semantic retrieval layer (ChromaDB) to ensure consistency and memory compaction.

• 🎯 Automated Brand Logo Extraction Zero-shot logo extraction pipeline combining Grounding DINO, SAM, and custom post-processing.
  Focus: Building a robust pipeline that operates without supervised training data.

I value the lessons learned from failed experiments as much as successful deployments. Below are R&D projects where I tested architectural hypotheses and made data-driven decisions.

• 🧪 Q-PSA (Project Killed) Tested discrete perturbation for quantized LLMs to estimate layer importance.
  Decision: Objectively killed the project after my data showed it was ~1300x slower than the baseline and failed pruning validations.

• 🗺️ Circle-WFC (Architectural Pivot) Attempted to replace A* pathfinding with a geometry-guided Wave Function Collapse (WFC). Insight: Discovered the structural limits of 'local consistency' in global pathfinding. Pivoted the concept's value into a highly efficient 'Search Space Reducer'.

• ⚡ HW-WFC v2.9 (Feasibility Validated) Constraint-driven AI compiler scheduling R&D.
  Result: Matched Exact DP's optimum, proving algorithmic feasibility, but strategically concluded the research after identifying hardware-backed cost-model calibration as the real production bottleneck.

• 🎓 ADIGA College Admission Data Pipeline Extracting and normalizing complex HTML data across 200+ institutions.
  Result so far: Reduced schema violation rates under noisy HTML inputs using hallucination-controlled LLM workflows.

For longer write-ups on troubleshooting, architectural decisions, trade-offs, and project context:
👉 Selected Project Details (PROJECTS.md)

Email: kdtyohan@gmail.com
LinkedIn: entangelk