• 

• 

AeroForge AI

Agentic AI Engineering Copilot for UAV design, manufacturing, procurement, risk, verification, and engineering traceability.

AeroForge AI transforms a structured engineering requirement set into a connected engineering decision workflow using a live AI model and specialist agent roles.

What it does

• AI Engineering Assistant — extracts measurable requirements, identifies unknowns, assumptions, and engineering considerations.
• Agentic Workflow — runs six specialist roles:
  1. Requirements Analyst
  2. Systems Engineer
  3. Aerospace Analyst
  4. Manufacturing Engineer
  5. Safety & Test Engineer
  6. Program Synthesizer
• BOM & Procurement — generates a starter engineering BOM and identifies procurement/specification gaps.
• Risk, Safety & Verification — generates an AI risk register and verification matrix.
• Engineering Decision Package — consolidates requirements, decisions, BOM actions, risks, verification, and open actions.
• Impact Dashboard — supports quantified comparison of manual versus AI-assisted engineering workflow time.

Demonstration scenario

The included demonstration uses a civil fixed-wing UAV engineering scenario for mapping/inspection:

• MTOW target: 10 kg
• Payload target: 2 kg
• Wingspan: 3.2 m
• Cruise speed: 80 km/h
• Endurance: 2 hours

All engineering outputs are screening-level decision support and require qualified engineering review before formal design release, procurement, or flight testing.

AI architecture

AeroForge AI uses an OpenAI-compatible client with Hugging Face Inference Providers. The demonstration configuration uses:

openai/gpt-oss-20b:fastest

The application also contains deterministic fallback behavior so the demonstration interface can remain usable when live inference is unavailable.

Run locally

Create a virtual environment and install dependencies:

python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -r requirements.txt

Create:

.streamlit/secrets.toml

with:

HF_TOKEN = "YOUR_HUGGING_FACE_TOKEN"
HF_MODEL = "openai/gpt-oss-20b:fastest"

Never commit secrets.toml or .venv to GitHub.

Start the application:

.\.venv\Scripts\python.exe -m streamlit run app.py

Hackathon workflow

Requirements → AI Analysis → Specialist Agents → BOM/Procurement → Risk → Verification → Decision Package → Impact

The objective is not to replace engineering judgment. AeroForge AI is designed to reduce repetitive analysis and documentation effort while keeping the engineer in the decision loop.

Safety and engineering responsibility

This project is intended for benign civil engineering applications such as mapping, inspection, environmental monitoring, research, and manufacturing support. AI-generated recommendations are advisory and must be independently validated by qualified engineers and against current drawings, datasheets, simulations, procedures, and applicable regulations.

Technology

• Python
• Streamlit
• Hugging Face Inference Providers
• OpenAI-compatible API
• Pandas
• PDF/DOCX/TXT/MD/CSV requirement ingestion

Project

AeroForge AI — AI Engineering Copilot

Built as a prototype for an AI productivity and innovation hackathon.

Copyright & Usage

© 2026 Engr. Aqsa Zehra / Skyward Innovators. All rights reserved.

AeroForge AI is an original project developed by Skyward Innovators as an Agentic Engineering Copilot for Civil UAV Programs.

This repository is provided for viewing, evaluation, demonstration, and educational purposes only.

Copying, modifying, redistributing, republishing, or presenting this project or substantial portions of its source code as another person's work is not permitted without prior written authorization.

For permission to use the source code, please contact the project authors.

Project Attribution

Project: AeroForge AI
Team: Skyward Innovators
Team Leader: Engr. Aqsa Zehra Team Member: Engr. Bilal Mehmood Domain: Avionics | UAV Engineering | Agentic AI
Year: 2026

© 2026 Engr. Aqsa Zehra / Skyward Innovators. All rights reserved.

CREATED BY ENGR. AQSA ZEHRA

Built With

• python