AI · Deep Learning · Computer Vision

Rakshit
Verma

B.Tech CSE at IIIT Dharwad.
Specialising in computer vision, state-space models, and biosignal processing.

Get in Touch GitHub ↗ LinkedIn ↗

/ skills

Technical Skills

Languages

PythonJavaC

Deep Learning

PyTorchTensorFlowHuggingFace Transformers

Computer Vision

OpenCVYOLOObject DetectionSemantic SegmentationHyperspectral Imaging

Architectures

State Space Models (Mamba)TransformersEncoder-Decoder CNNs

ML Utilities

NumPyPandasScikit-learnMatplotlibSeaborn

Backend

DjangoFastAPIREST APIsSQLiteDocker

Tooling

GitGitHub ActionsJupyterLinuxGoogle Colab

/ research

Research

Hyperspectral Image Generation using Spectral-Spatial State-Space Modeling

Under Review
  • Designed generative architecture for hyperspectral reconstruction using spectral-spatial fusion.
  • Integrated state-space modeling to capture long-range spectral dependencies.
  • Built end-to-end PyTorch training pipeline for multi-band tensor processing.
  • Achieved state-of-the-art results on the CAVE benchmark dataset (PSNR, SSIM).

EMG Signal Classification for Prosthetic Control using Mamba

Publication in Progress
  • Implemented Mamba-based state-space architecture for long-range temporal modeling of EMG signals.
  • Developed preprocessing pipeline for multi-channel biosignal normalization.
  • Achieved improved classification performance over LSTM and GRU baselines.
  • Application: real-time prosthetic limb control.

/ projects

Projects

Constellation Detection System

Live ↗
Hackathon · Round 3
YOLODjangoGPSAI Chatbot
  • Trained 16-class object detector — mAP@50 95.8%, Precision 90.5%, Recall 95.6%.
  • Integrated real-time inference into Django backend with GPS-based sky recommendations.
  • Developed multilingual AI chatbot with voice input/output.

Multimodal Representation Learning with CLIP

Amazon ML Challenge · Top 300
CLIPPyTorchEmbeddings
  • Leveraged pretrained CLIP encoders for embedding-based multimodal modeling.
  • Built scalable PyTorch inference pipeline for large-scale feature extraction.

Semantic Segmentation using U-Net

TensorFlowU-NetSegmentation
  • Implemented encoder-decoder U-Net (8.6M parameters) for multi-class segmentation.
  • Built tf.data preprocessing pipeline; trained using Sparse Categorical Crossentropy.

/ certifications

Certifications

Oracle ✓ Certified Professional

Oracle Certified Professional — Generative AI

DeepLearning.AI

Convolutional Neural Networks

DeepLearning.AI

Neural Networks and Deep Learning

University of Michigan

Building Web Applications in Django

NIELIT

Data Analysis with Python

/ achievements

Achievements

1st

Developed a real-time edge detection model deployed on Raspberry Pi, trained on MNIST — demonstrated live embedded inference at the competition.

2nd

Built a neural network from scratch in C with no ML frameworks — implementing forward pass, backpropagation, and gradient descent at the systems level.

/ contact

Get in Touch

Send me an email

Open your mail client with a pre-filled message — just fill in your details and hit send.

✉ Open Email Draft ↗
GitHub github.com/rakshverma ↗ LinkedIn linkedin.com/in/rakshit-verma ↗