Top 10 research topics in AI for BTech Final year Project

For a final year BTech project in 2026, research-oriented topics should move beyond basic "classification" and focus on Generative AI, Trustworthy AI, and Edge Intelligence. These areas demonstrate high technical maturity and are currently high-priority in both industry and academia.

Here are the top 10 research-oriented AI project topics:

1. Multimodal Generative AI for Medical Report Synthesis

Instead of just classifying an X-ray, this project uses a Multimodal Large Language Model (MLLM) to analyze an image and generate a structured medical report.

• Research Focus: Reducing "hallucinations" in AI-generated medical text and improving cross-modal alignment between visual pixels and medical terminology.

• Tech Stack: PyTorch, Hugging Face Transformers, LLaVA or CLIP models.

2. Federated Learning for Privacy-Preserving Healthcare

Train a global AI model on patient data across multiple hospitals without the data ever leaving the local servers. This addresses the critical research challenge of data privacy.

• Research Focus: Optimizing communication overhead between clients and the server, and handling non-IID (Independent and Identically Distributed) data.

3. Explainable AI (XAI) for Credit Scoring Models

Financial institutions are hesitant to use "black box" AI. This project builds a credit risk model that provides local and global explanations for why a loan was approved or rejected using SHAP or LIME.

• Research Focus: Comparing the fidelity of different explanation methods and their impact on user trust.

4. Real-Time Deepfake Detection in Video Calls

With the rise of sophisticated "AI-swapped" faces, research into detecting temporal inconsistencies in live video streams is highly relevant.

• Research Focus: Identifying physiological inconsistencies (like blink patterns or pulse detection via skin color changes) that Generative Adversarial Networks (GANs) often miss.

5. Efficient LLM Fine-Tuning for Low-Resource Languages

Most LLMs work poorly for regional dialects. This project uses QLoRA or LoRA (Low-Rank Adaptation) to fine-tune a model on a specific local language (e.g., Marathi, Telugu, or Swahili) with limited hardware.

• Research Focus: Measuring the "catastrophic forgetting" of the base model when fine-tuned on a tiny, specific dataset.

6. AI-Driven Resource Allocation in 6G Networks

Use Reinforcement Learning (RL) to predict network traffic and dynamically allocate bandwidth or power in a future 6G environment.

• Research Focus: Implementing "Deep Q-Learning" to minimize latency in high-density IoT environments.

7. Autonomous Drone Navigation using Vision Transformers (ViT)

Replace traditional CNNs with Vision Transformers for drone navigation to provide better global context of the environment, helping the drone navigate complex, unstructured spaces.

• Research Focus: Optimizing Transformer models for real-time inference on edge devices like Jetson Nano.

8. AI for Carbon Footprint Optimization in Smart Buildings

Develop an AI agent that controls HVAC and lighting systems based on real-time occupancy and electricity price fluctuations to minimize carbon emissions.

• Research Focus: Multi-objective optimization (balancing comfort vs. energy saving) using Genetic Algorithms or PPO (Proximal Policy Optimization).

9. Adversarial Attacks and Defenses on Computer Vision

Research how "adversarial noise"—invisible to humans—can fool an AI into misidentifying objects (e.g., making a stop sign look like a 60mph sign).

• Research Focus: Developing "Adversarial Training" techniques to make models robust against these security threats.

10. Graph Neural Networks (GNN) for Fake News Detection

Instead of looking only at the text, analyze the propagation graph—how the news spreads on social media—to identify bots and misinformation clusters.

• Research Focus: Capturing structural patterns of misinformation spread using Graph Convolutional Networks (GCNs).

Project Feasibility Matrix