Hello!

I'm Soham Pahari, a final-year Computer Science (Data Science) student specializing in machine learning, deep learning, and applied AI research. I work across multimodal learning, vision transformers, personality modeling, drug discovery frameworks, and optimization-driven forecasting models.

I have industry and research experience through roles at Bahas Pvt Ltd and NIT Warangal (Dept. of AI), where I led and contributed to projects in emotion classification, molecular sequence generation, and high-performance model training on large-scale GPU systems. My work spans transformer-based architectures, multimodal reasoning, and scalable ML engineering.

Some of my recent research highlights include:

• Developing AMoE-ViT, an Adaptive Mixture of Orthogonal Experts Vision Transformer for self-supervised agricultural disease diagnosis.
• TriadNet, a multimodal system integrating graph-based facial dynamics, prosody-aware speech modeling, and BERT-LSTM textual flow to estimate Big Five personality traits.
• Leading a 1.13M-sample drug–indication generation project as a research intern and achieving strong improvements over baseline molecular benchmarks.
• Building multiple ML solutions including AQI forecasting models, interview simulation systems, and emotion recognition applications.

I’ve presented my work at conferences such as COMSYS 2025 (Warsaw University of Technology), KEDAR 2025, and CIACON 2025, and I am also the recipient of the SHODH Research Grant for an IoT-driven agricultural system.

Outside of research, I enjoy working on end-to-end systems, exploring applied AI for real-world impact, and competing in hackathons—recently placing top 200 out of 17,000 participants at Hackahazard 2025 and winning the Smart India Hackathon internal round.

If you're interested in machine learning research, multimodal AI, or applied deep learning engineering, I’d love to connect.