Professional Photo
(180x180px)

Zhigang Tian (Hanson Tins)

AI Neuroscience Researcher & Principal Engineer
Pioneering research at the intersection of neuroscience and artificial intelligence, developing novel architectures inspired by brain mechanisms for enhanced AI safety and performance. Founder of multiple AI ventures with expertise in multi-agent systems, medical AI, and neuromorphic computing.

AI Neuroscience Research Framework

Systematic research agenda based on amygdala hijacking experiments and neuroscience-inspired AI systems

Phenomena Level

Discover and quantify neuron-like phenomena in AI

Structural Level

Understand architectural components' impact

Dynamics Level

Study temporal evolution and dynamic properties

Statistical Level

Establish universal laws and distributions

Defense Level

Develop neuroscience-based safety mechanisms

Module 1: AI Emotion Modeling & Hijacking

  • Induced hijacking critical thresholds
  • Spontaneous hijacking noise-memory phase diagrams
  • Gating thresholds and feedforward inhibition
  • Emotional attractor dynamics
  • Power-law duration distributions

Module 2: AI Inhibition Mechanisms

  • Feedforward inhibition design
  • Feedback inhibition mechanisms
  • Network stability enhancement
  • Adaptive inhibition regulation

Module 3: AI Memory & Consolidation

  • Dual-stage memory architecture
  • SPW-R inspired replay mechanisms
  • Selective memory consolidation
  • Energy-constrained memory systems

Module 4: Learning & Plasticity Dynamics

  • Bidirectional plasticity rules (LTP/LTD)
  • Adaptive learning rate mechanisms
  • Anomalous learning pattern detection
  • Bias propagation dynamics

Module 5: Sequence Modeling & Structure

  • θ-γ coupling simulation
  • Hierarchical temporal modeling
  • Neural grammar constraints
  • Oscillatory hierarchical structures

Featured Projects

AgentForest.one

Multi-Agent SaaS Platform

Built a multi-agent SaaS platform orchestrating domain agents via shared Model Context Protocol (MCP) with full-stack deployment.

Fly.io Neon Postgres Cloudflare R2 RunPod/vLLM MCP

MedGPT.co

Medical AI Platform

AI-based medical toolset with 3,000+ organic users, integrating Hugging Face, OpenAI APIs, and Med-PaLM models for comprehensive medical solutions.

Node.js Express OpenAI API Med-PaLM Flutter

P-adic Machine Learning

Mathematical AI Research

Novel approach to machine learning using p-adic number theory for enhanced numerical stability and pattern recognition.

Python NumPy TensorFlow p-adic Analysis

ESR-GNN

Graph Neural Networks

Enhanced spectral residual graph neural networks for improved representation learning and node classification tasks.

PyTorch PyG NetworkX Spectral Analysis

Professional Experience

2024 — Present

Founder & Principal Engineer

AgentForest.one | Remote

Leading end-to-end architecture & implementation of multi-tenant SaaS platform with MCP-based shared context and agent orchestration. Implemented AI serving with OpenAI/LLaMA APIs, RAG pipelines, and dynamic model routing.

2023 — Present

Founder & Lead Engineer

MedGPT | Philippines

Led technical architecture of MedGPT.co with 3,000+ organic users. Managed development of Android version and AI talking doctor platform using Flutter 3.7 and Python.

2021 — 2022

Cloud Engineer

Accenture | Philippines

Standardized IaC (Terraform) and container delivery (Docker/K8s) across GCP/Azure/AWS environments. Implemented CI/CD with versioning, canary deployments, and comprehensive monitoring.

2019 — Present

Founder & Project Supervisor

FacePay & FaceBanc | Philippines

Developed unified payment platform supporting major payment systems across Southeast Asia, Middle East, Europe, and North America with one-click enterprise access.

Achievements & Recognition

🏆

Patent Application

Type/No: Invention 1/2023/050166

Title: GPT-based Automated Medical Care System

📜

Software Copyrights

12 Registered Copyrights

Including MedGPT AI Doctor, MedCloud.PH Platform, and various medical AI systems

🎓

Education

Bachelor's in ML & AI

Goldsmiths, University of London (2021-2024)

👥

User Impact

3,000+ Users

Organic growth of MedGPT platform with 100+ daily new users

Research Publications & Roadmap

Planned Publication Strategy

Phase 1: Phenomena Discovery & Modeling (Papers 1-4)
Target Venues: NeurIPS, ICML, ICLR
Focus: Induced hijacking thresholds, spontaneous hijacking phase diagrams, gating mechanisms, emotional attractor dynamics
Phase 2: Statistical Laws & Universality (Papers 5-6)
Target Venues: Nature Machine Intelligence, Science Robotics
Focus: Power-law duration analysis, information bottleneck β-phase transitions
Phase 3: Architecture Optimization & Applications (Papers 7-8)
Target Venues: AAAI, Neural Networks
Focus: MEGA component causal analysis, defense mechanism minimal sets
Phase 4: Neuroscience-Inspired Extensions (Papers 9-16)
Target Venues: Cognitive Science, Connection Science
Focus: Deep research based on five research modules