Ricardo portrait

Hey, Saif here. Crafting AI dreams into reality — one neural network at a time.

My projects

GPTX

    Made with:

Built a 10.8M-parameter Bigram LM with character embeddings, achieving 1.07/1.48 train-val loss over 5K CUDA iterations. Engineered an 8-token context model with 32-token batches for 501-token Shakespearean text generation via softmax sampling. Optimized training efficiency with PyTorch, leveraging GPU acceleration for faster convergence and improved text coherence.

AMORA - save your love

    Made with:

An AI-powered gift-sending platform with automated reminders via SMS, WhatsApp, and email. Designed a modern, visually rich UI using Next.js and Tailwind CSS. Integrated Supabase for backend management and AI-driven personalization to enhance user experience.

Magic Calc

    Made with:

Developed a web-based tool that visualizes mathematical problems and provides solutions. Utilized FastAPI and Gemini API for dynamic solution visualization, integrated with a Vite and React frontend. Enhanced user interaction with real-time rendering and AI-powered explanations for a seamless learning experience.

DeepFake Detection

    Made with:

Developing 'Analyze the Temporal and Spatial Inconsistencies in Deepfake Videos Using Deep Learning Architectures', leveraging CNNs and RNNs to detect deepfake artifacts. Focused on enhancing video forensics by analyzing temporal frame inconsistencies and spatial anomalies for robust deepfake detection

Anomaly Detection for industrial equipment

    Made with:

Developed an industrial anomaly detection system leveraging Anomalib, PatchCore, and ResNet to automate defect detection in manufacturing. Utilized deep learning-based feature extraction and patch-based anomaly localization to improve inspection accuracy. Enhanced quality control by minimizing manual efforts and increasing detection precision

F1 predictor

    Made with:

Developed an F1 race winner prediction model using machine learning. Analyzed historical race data, driver performance, track conditions, and team strategies to enhance predictive accuracy.

BERT Sentiment Analysis

    Made with:

Implemented a sentiment analysis model using BERT to classify text with high accuracy. Fine-tuned transformer-based embeddings on domain-specific data to enhance sentiment detection and contextual understanding

My Skills

  • Python
  • Deep Learning
  • AI/ML
  • Pytorch
  • TypeScript
  • Tensorflow
  • Hugging Face
  • Django
  • Numpy
  • Java
  • React
  • Javascript
  • Node.js
  • Pandas
  • FastAPI
  • C++
  • Next.js
  • Prompt Eng
  • tailwind
  • Hadoop
  • Git
  • MySQL
  • PostgreSQL
  • Linux
  • AWS
  • Docker

My experience

My Education

QUT

Bachelor of Technology Computer Science

Marwadi University (NAAC A+), Rajkot, India

Expected Graduation in 2026

GPA 8.5/10

  • Tech Lead at TECHBRIDGE community
  • Active participant in hackathons and coding competitions
  • Member of Green Club MU and involved in social outreach programs

Contact me

Please contact me directly at saifr1497@gmail.com or through this form.