dijju

TRUE / FALSE:

Q1. A vague prompt usually gives a precise and useful answer.
 Answer: False

Q2. Adding constraints like word limit improves AI responses.
 Answer: True

Q3. AI always provides factually correct answers.
 Answer: False

dijju

Exercise : Team Challenge Round aligned with your chosen domains:

 1. Climate
 2. Composite Materials
 3. Smart Grid
 4. Autonomous Vehicles
 
 
 Challenge Setup :  Core Task

dijju

Trainer Bridge

“Even with a good prompt… AI rarely gets it perfect in one shot.”

“The real skill is not writing one prompt —
it’s improving it over multiple turns.”



Core Concept (Explain Quickly)

dijju

Prompt Techniques — Role, Delimiters & Step-by-Step

Trainer Bridge

“Now you know how to structure a prompt…
Let’s make it smarter.”

“Same prompt — but with small techniques — can give dramatically better results.”

Step 1: Demo — Same Prompt, Better Output (5 min)

dijju

GOAL:
“The goal is simple:
Turn a weak prompt into something AI can actually work with.”


Step 1: Show the Problem
Display 3 Real Prompts (from Exercise 1)
  “Tell me about AI”
  “Explain photosynthesis”
  “Help me with coding”

dijju

1. “Before we learn how to talk to AI, I want to see how you already think about it.”

REMEMBER: “There are no wrong questions — only interesting ones.”

2. Instruction:

“Think of ONE thing you would ask an AI tool”


Constraint:

dijju

The following deliverables are expected to be completed by end of the day .. by choosing the project title from the following list:
AGRICULTURE:
1. Enable on-field, real-time detection of early signs of stress in crops using efficient image pipelines and deployable models without cloud dependency.
2. Automate weed detection in the field using computationally inexpensive CNNs to reduce herbicide usage and increase precision agriculture yield.
3. Build a lightweight, transformer-based model that can recognize plant growth stages on-device without internet, enabling agronomists to track phenological transitions for yield forecasting and nutrient planning.
HEALTHCARE:
1. Provide affordable, contactless respiratory health screening on edge devices, especially for rural clinics or public entry points.
2.Enable real-time cancer screening and triaging in low-resource hospitals using a privacy-preserving, deployable edge AI system.
3. Create an accurate, low-cost, non-contact method to monitor respiration for critical

dijju

# main.py
# Full edge-camera pipeline orchestration
# - Capture from webcam
# - Sample frames at controlled FPS
# - Preprocess each frame
# - Simulate model inference
# - Track performance

from camera.webcam import Webcam
from pipeline.sampler import FrameSampler

dijju

# metrics.py
# Responsibility:
# - Track frames per second (FPS)
# - Track memory used (MB)

import time
import psutil


class Monitor:

dijju

# dummy_model.py
# Responsibility:
# - Simulate the latency and structure of a real model
# - Let us test the end-to-end pipeline without ML distractions

import time
import numpy as np


def dummy_inference(frame):