Aditya Tanna

Email: [email protected] LinkedIn: www.linkedin.com/in/aditya-tanna29

Contact: +91 76004 43472
Education
Dhirubhai Ambani Institute of Information and Communication Technology Gandhinagar, India
Bachelor of Technology - Mathematics and Computing; 7.92/10 (till Semester 7) August 2021 - Present
Delhi Public School - Modern Indian School Doha, Qatar
Class 10 and 12 - CBSE July 2018 - May 2021
– Secured 94.4% in Class 10 and 95.8% in Class 12 Board Exams.
Skills Summary
• Area of Interests: Machine Learning, Natural Language Processing (Personalization & Summarization), Large Language
Models
• Tools/Languages: LangChain, Spacy, PyTorch, TensorFlow, Numpy, Matplotlib, Scikit-learn, Pandas, Neo4j, FastAPI,
Jupyter Lab, MySQL, Spark, MongoDB, DynamoDB, Docker, OpenHands
• Electives: Database Management Systems, Operating Systems, Data Structures and Algorithms, Information Retrieval,
Natural Language Processing, Big Data Processing

Experience
Knowledge and Discovery Lab - KDM Lab Gandhinagar, India
Research Assistant October 2023 - Present
– Leading the development of PerDucer, a personalized summarizer leveraging LSTM and attention mechanisms
for user-specific summaries, based on their history sequence.
– Conducted an in-depth study on Temporal Knowledge Graphs for capturing evolving user preferences along
with Explainable AI/LLMs; developed a novel data augmentation framework (PerAugy) leveraging Double
Shuffling and Stochastic Markovian Perturbation, submitted to ACL’25.
Teaching Assistant (TA) Gandhinagar, India
IE494 - Big Data Processing ; MC221 - Database Management Systems (DBMS) July 2024 - December 2024
– Assisting in assignments and grading for the Big Data Processing and DBMS courses.
– Collaborating with professors to help students understand complex concepts related to both courses.
Projects
FactRAG - Automated Fact Checking using GraphRAG September 2024 - Present
Skills: LangChain, Information Retrieval, Neo4j, Knowledge Graphs,RAG
– Implemented a graph-native design using Neo4j and Cypher for efficient knowledge graph querying, supporting
multi-hop fact verification with up to 2-node traversal.
– A scalable system architecture integrating the processing layer, Llama 3.1.8b model, and graph retriever for
handling the Wikidata5M dataset.
PerAugy - Perturbation-based Augmentation (Submitted for ACL -25) June 2024 - September 2024
Skills: - Data Augmentation, LLM - Data Generation, User Preference Modeling Professor Sourish Dasgupta
– Developed PerAugy,in collaboration with LCS2 Lab @ IIT-Delhi a cross-trajectory data augmentation
technique to tackle limited diversity in user preference data for personalized summarization.
– Incorporated Double Shuffling and Stochastic Markovian Perturbation (SMP) to synthesize realistic user
interactions, capturing multi-aspect preferences.
– Demonstrated significant boosts in dataset diversity, user-encoder accuracy, and overall personalization metrics,
as evidenced by empirical results on the PENS dataset.
PerDucer - Personalized Summarizer August 2024 - Present
Skills: Temporal Knowledge Graphs, Pointer-Generator Networks Professor Sourish Dasgupta
– Developed PerDucer, a personalized summarization model using Temporal Knowledge Graphs to capture
evolving user preferences. Implemented a dual encoder architecture for dynamic user behavior modeling and
preference adaptation.
Weather Forecasting - Time Series Analysis April 2024 - July 2024
Skills: Time Series Analysis, Machine Learning, Statistics
– Modeled temperature and humidity using SARIMAX to account for seasonal and non-seasonal patterns.
Analyzed stationarity, seasonality, and rolling averages to uncover long-term trends of three cities, Jerusalem,
Los Angeles and Miami.
Awards
• Awarded Merit scholarships from DAIICT for academic excellence (Semesters 1 through 7).
• Academic Excellence Award from DPS-MIS for Classes 9 through 12.