kendrickfff/ai-pm-ken-advisor

🧠 Ken Advisor — Fine-Tuned Phi-2 for Product Management Assistance

Model name: ken-fine-tuned-phi2

Base model: microsoft/phi-2

Architecture: Decoder-only Transformer (Causal LM)

Fine-tuning framework: LoRA (Low-Rank Adaptation) with PEFT

Dataset type: Instruction-style chat pairs for Product Management scenarios

---

📘 Model Overview

Ken Advisor is a fine-tuned version of Microsoft’s Phi-2, trained to serve as an AI Product Management Assistant. It provides structured, concise, and context-aware guidance across various product management tasks — from ideation and discovery to stakeholder communication and AI integration.

This model was trained using a LoRA adaptation, allowing efficient fine-tuning on consumer GPUs while preserving Phi-2’s original reasoning strength.

---

💡 Key Features

• Responds to PM-related queries naturally and informatively
• Understands user-product conversation flows
• Produces actionable insights for product strategy, UX, and AI integration
• Lightweight (2.7B parameters) and efficient for deployment

---

🧩 Training Details

Parameter	Value
Base model	microsoft/phi-2
Method	LoRA fine-tuning (PEFT)
Dataset size	~540 samples
Epochs	3
Batch size	1
Learning rate	2e-4
Max sequence length	512
Average training loss	0.2468
Framework	Hugging Face Transformers (v4.38)

---

🧠 Example Usage

from transformers import pipeline

pipe = pipeline("text-generation", model="your-username/ken-fine-tuned-phi2")

prompt = "User: How should I prioritize AI features in my product roadmap?\nAssistant:"
result = pipe(prompt, max_new_tokens=150)
print(result[0]["generated_text"])

---

⚙️ Intended Use

This model is designed for:

• Product managers and startup founders
• AI-driven project teams
• PM education and mentorship use cases

⚠️ Limitations: Not suitable for legal, financial, or medical advice. Outputs may include hallucinations or inaccuracies — always verify before use.

---

🏷️ Citation

If you use this model, please cite:

@model{ken_advisor_phi2,
  title={Ken Advisor — Fine-Tuned Phi-2 for Product Management},
  author={Kendrick F.},
  year={2025},
  base_model={microsoft/phi-2},
  framework={Transformers + PEFT}
}

---