AADITYA SINGH TARIYAL

About

AI/ML Engineer specializing in Graph-RAG architectures and multilingual NLP, with experience in automating legal analysis and predictive modeling. Built and deployed end-to-end AI solutions spanning computer vision, forecasting, and retrieval-augmented workflows.

Experience

• AI/ML Intern
  Google for Developers – AICTE EduSkills
  Jan 2026 – Mar 2026
• Artificial Intelligence Intern
  AICTE SkillsBuild
  Jul 2025 – Jul 2025
• Data Science Intern
  CertED Technologies
  Jun 2025 – Jun 2025

Education

• B. Tech Computer Science and Engineering
  Haridwar University
  Computer Science and Engineering · Sep 2023 – Jun 2027

Skills

• SQL
• FastAPI
• Git
• GitHub
• Python
• Jupyter Notebook
• Power BI
• MySQL
• JavaScript
• MongoDB
• Django
• Matplotlib
• NumPy
• Pandas
• scikit-learn
• OpenCV
• PyTorch
• TensorFlow
• Docker
• LangChain
• LangGraph
• Faiss
• Chromadb
• Ibm Watson

Projects

• Spotify Music Suggestion System ↗
  Python, Scikit-Learn, PCA, Streamlit

  Performed EDA, data cleaning, and feature scaling, followed by PCA and K-Means clustering for user preference segmentation. Built a content-based music suggestion system using Streamlit allowing users to input a song or artist and receive similar track suggestions.

• Multilingual Hate Text Detection ↗
  MuRIL, Transformers, Gradio

  Fine-tuned MuRIL for hate speech classification across Hindi, English, and Hinglish using a 29K+ multilingual dataset with custom preprocessing and label validation. Iterated on MuRIL architecture to improve F1-score to 0.75, outperforming baseline models by 12%. Deployed an interactive web demo on Hugging Face Spaces with Gradio for real-time multilingual text classification and confidence scoring.

• AI Legal Advisory Assistant ↗
  Graph-Based RAG, LangGraph, FastAPI, LangChain, ChromaDB

  Built a FastAPI service powered by a LangGraph-based agentic RAG workflow to automate contract clause extraction and legal intent validation, reducing manual review time by 75%. Designed a metadata-aware retrieval pipeline using LangChain and ChromaDB to achieve 95% traceability with filename and page-level citations, minimizing hallucinations in legal responses. Optimized token usage by 40% through prompt engineering and guardrail implementation, reducing operational costs.

Courses & certifications

• Getting Started with Artificial Intelligence · IBM · 2026
• Machine Learning Training Course · IntrnForte · 2025

🗣️ Languages

• English · Intermediate
• Hindi · Native

🎯 Hobbies & interests

• Chess
• travelling
• brainstorming startup ideas