Location: Noida, Uttar Pradesh, India
Contact: 251shreesh@gmail.com
GitHub: github.com/cosmoEagle
LinkedIn: linkedin.com/in/shreesh-shukla
I am an AI and Machine Learning enthusiast pursuing B.Tech in Computer Science Engineering at Bennett University. With a deep interest in large language models (LLMs), transformer architectures, and legal tech, I focus on solving real-world problems through AI. My expertise in Python, GANs, transformers, and prompt engineering allows me to design impactful solutions.
-
Bachelor of Technology (Computer Science & Engineering)
Bennett University (2022 – Present)
CGPA: 9.90/10 -
Class XII (CBSE Board)
Jawahar Navodaya Vidyalaya (2020–2021)
Grade: 96.40% -
Class X (CBSE Board)
Jawahar Navodaya Vidyalaya (2018–2019)
Grade: 96.40%
IITI Drishti Foundation (Oct 2023 – Mar 2024)
- Developed AI tools for legal research, focusing on accident law analysis.
- Conducted prompt engineering and utilized LLMs to increase AI accuracy by 30%.
- Collaborated with multi-functional teams, integrating AI solutions to improve research efficiency.
- Designed an LLM-powered tool to streamline bail applications by analyzing chargesheets.
- Delivered legally compliant outputs with 65–70% efficiency compared to manual findings.
- Created an AI-powered solution to simplify legal jargon and automate document drafting.
- Utilized LLaMAIndex, LangChain, and LLMs to streamline processes, reducing legal documen 709D tation time by 70%.
- Technologies: LLMs, LangChain, LLaMAIndex, NLP, Transformer Models
- Developed a voice-controlled assistant for automated logins and Wi-Fi management using Python and Selenium.
- Achieved 95% accuracy in recognizing and managing user login sessions.
- Technologies: Selenium, Voice Recognition, Python, Database Management
A.Saran, S. Shukla, and T. Ahmed et al. "A comparative analysis of zero-shot rhetorical role classification in the legal domain"
International Conference on AI and the Digital Economy (CADE 2024), Venice, Italy, 2024.
DOI: 10.1049/icp.2024.2542
- Explored the application of zero-shot learning for classifying rhetorical roles in legal documents.
- Compared performance across various transformer models, with DeBERTa achieving the highest F1-score of 0.7497.
- Python, C++, Java, SQL, JavaScript
- LangChain, PyTorch, TensorFlow, LLaMAIndex
- Git, Android Studio, Microsoft SQL Server