Dinesh S.

Key Skills

Software

Other

Top Subject Matter

Brain imaging reconstruction (fMRI) and text-conditioned generative modeling

DAI (distribution fitting) with Gaussian mixture modeling

Text similarity and plagiarism detection algorithms

Top Data Types

Text

Image

Top Task Types

Prompt + Response Writing (SFT)

Text Generation

Freelancer Overview

Brain Image Reconstructor | Course Project (AI and ML) (April 2025). Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other. Education includes Bachelor of Technology, Indian Institute of Technology Bombay (2023) and Minor Program, Indian Institute of Technology Bombay (2023). AI-training focus includes data types such as Medical, DICOM, and Computer Code and labeling workflows including Prompt + Response Writing (SFT), Computer Programming, and Coding.

Labeling Experience

Brain Image Reconstructor | Course Project (AI and ML) (April 2025)

OtherPrompt + Response Writing (SFT)

Worked on an AI/ML project extending an existing brain image reconstruction approach by altering the fMRI-to-text conditioning mechanism using user prompts. Built a two-stage generative pipeline that uses latent-variable modeling and diffusion with CLIP guidance, and varied diffusion/control parameters to study output changes. Evaluated how prompt-driven conditioning and mixing strengths affected model outputs and reconstruction quality. • Replaced an fMRI-to-text regressor with user prompt conditioning. • Integrated CLIP-feature guidance into diffusion-based generation. • Added controls for diffusion and mixing strengths for experimentation. • Conducted output evaluations across parameter variations.

2025

Face Recognition Using Siamese Networks | Self Project (Summer 2024)

OtherTextText Generation

Developed a face recognition system using Siamese networks that learns similarity between images. Trained twin neural networks with shared weights and a contrastive objective to distinguish similar versus dissimilar faces. Used a triplet-loss setup with an L1 distance layer to improve separation for closely related identities. • Implemented Siamese/twin networks with shared weights. • Used contrastive loss for learning facial similarity. • Applied triplet loss with L1 distance to separate similar faces. • Trained the model to discriminate closely related individuals.

2024

Plagiarism Detector | Course Project: Data Structures & Algorithms Lab (Autumn 2024)

Other

Created a plagiarism detection system implementation for a data-structure lab project. The work focused on efficiently comparing candidate texts using matrix-based techniques and reducing false positives/negatives through careful analysis. It also included security testing by attempting to bypass vulnerabilities in peer systems. • Implemented a C++ plagiarism checker with matrix-based matching. • Built a multi-threaded/asynchronous pipeline for real-time bulk processing. • Performed timestamp-based analysis to minimize errors. • Conducted security testing by crafting bypass exploits.

2024

Gaussian Mixture Model for Distribution Fitting | Course Project: DAI (Autumn 2024)

Other

Developed a probabilistic distribution-fitting model as part of an AI coursework project. Although primarily a modeling task, it involved creating and validating a statistical fitting pipeline used to represent data distributions accurately. Produced visualizations to assess the fit quality of the learned mixture model. • Implemented a 3-component Gaussian Mixture Model in Python. • Used optimization (fsolve/min negative log-likelihood) to fit parameters. • Generated histograms and probability density plots for model assessment. • Applied numerical methods to ensure accurate distribution modeling.

2024

Education

Work History