Skip to content
View Tarak-Ram-1908's full-sized avatar

Highlights

  • Pro

Block or report Tarak-Ram-1908

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Tarak-Ram-1908/README.md

Alahari Tarak Ram

B.Tech in Computer Science & Engineering | IIITDM Kancheepuram Aspiring AI Research Engineer | Focus: Computer Vision, Multimodal ML, Natural Language Processing


🚀 About Me

I am a 5th-semester Computer Science student at IIITDM Kancheepuram (CGPA: 8.67) dedicated to advancing the state of the art in Artificial Intelligence. I am actively building a research portfolio aimed at top-tier MS programs and global R&D roles.

  • 🔭 Current Focus: Multimodal Visual Question Answering for sports video understanding — fine-tuning LLaVA-1.5-7B with LoRA on a custom-annotated Cricket VQA dataset.

🧬 Selected Research & Projects

Building a domain-specific multimodal VQA system for T20 broadcast footage. Involves constructing a custom annotated dataset of 500+ frames across five question types (counting, spatial, formation, context, tactical delivery prediction), and fine-tuning LLaVA-1.5-7B with LoRA adapters.

End-to-end LightGBM win-prediction system trained on 703 Cricsheet matches. 63 engineered features across ELO ratings, H2H history, venue familiarity, phase-by-phase batting/bowling ratings, and toss dynamics. Achieved 93.5% accuracy on held-out 2022 & 2024 World Cup data. Deployed as a GitHub Pages site.

Fine-tuning IndicBART for low-resource Telugu→English NMT using RAMP data augmentation on the Samanantar corpus. Exploring back-translation and transfer learning to improve BLEU scores.

Schema-aware Text-to-SQL system fine-tuned on the Spider dataset. Focused on domain adaptation and improving generalization of T5-small to complex multi-table queries.


🛠️ Technical Skills

  • Languages: Python, C++, C, Verilog, SQL
  • AI/ML Frameworks: PyTorch, Hugging Face Transformers, Scikit-learn, LightGBM
  • Libraries: OpenCV, NumPy, Pandas, Matplotlib, Seaborn
  • Research Interests: Multimodal Machine Learning, Computer Vision, Natural Language Processing, Reinforcement Learning, Time Series Forecasting

📫 Connect with Me


"Precision in code, innovation in research."

Popular repositories Loading

  1. Tarak-Ram-1908 Tarak-Ram-1908 Public

    1

  2. text2sql-transformer text2sql-transformer Public

    Schema-aware Text-to-SQL Transformer

    Python 1

  3. Telugu-to-English-NMT Telugu-to-English-NMT Public

    Fine-tuning IndicBART for low-resource Telugu→English NMT using RAMP augmentation on the Samanantar corpus. Epoch 1/5 complete.

    Python 1

  4. OLD-Projects OLD-Projects Public

    This repository contains a collection of my old mini Python projects built while practicing programming.

    Python

  5. FoodBridge FoodBridge Public

    Real-time surplus food redistribution platform with ML-based spoilage prediction and volunteer route batching

    JavaScript

  6. wc2026-predictions wc2026-predictions Public

    ICC Women's T20 World Cup 2026 — ML Predictions

    HTML