7 Core Building Blocks of AI Agents
- Developed a modular framework to experiment with autonomous AI agents capable of reasoning, planning, and tool use.
- Implemented multi-agent collaboration using LangGraph and LangChain for dynamic role assignment and message passing.
- Integrated LLM-driven workflows (task planning, execution, and feedback loops) to demonstrate real-world agentic behavior in domains like grocery planning and tutoring.
This project conducts a deep exploratory data analysis (EDA) on the Rent the Runway dataset, exploring user reviews, product metadata, customer demographics, and seasonal trends to derive actionable business insights and hypotheses.
- Performed user behavior & demographic analysis, revealing rental trends by age group, body type, and frequency, plus insights into core customer segments.
- Analyzed product categories and fit feedback, identifying which apparel types are rented most often and how sizing issues affect review sentiment.
- Explored temporal & sentiment trends, such as seasonal peaks, review sentiment distributions, and relationships between review length and rating extremes.
This project studies the impact of a policy change implemented by OpenSea in January 2022. It leverages the OpenSea API to collect and analyze data on creators and their collections (before vs. after policy), along with asset activity levels to observe how changes affected creator behavior and asset dynamics.
- Retrieved creator and collection metadata (for creators who joined before Jan 2022) to compare collection growth, asset counts, and project diversity.
- Gathered and analyzed asset-level activity (sales, transfers, listings) to measure shifts in market engagement and liquidity post-policy change.
- Produced insights on platform-wide trends—e.g. whether policy adjustments led to consolidation among creators, changes in top-performing collections, or shifts in volume distribution.
Reusable Python plotting utilities for EDA & data visualization.
- Built wrappers for line, bar, scatter, and heatmaps
- Reduces repetitive plotting code by ~40%
- Tech: Python, Pandas, Matplotlib, Seaborn
Implementing core algorithms and data structures from first principles.
- Includes sorting, searching, trees, graphs, and ML basics.
- Clear explanations + complexity analysis.
- Tech: Python, NumPy
Hands-on Natural Language Processing coursework.
- Covers preprocessing, embeddings, tagging, parsing, evaluation .
- Experiments and reproducible reports included .
- Tech: Python, NLP libraries
This project explores classifying tweets into categories (e.g. sentiment, topic, or other labels) using classical ML models and transformer-based approaches. It includes data preprocessing, feature extraction, model training, evaluation, and comparison of performance across methods.
- Data preprocessing & feature engineering: Tokenization, stopword removal, TF-IDF, embedding-based features, and handling class imbalance.
- Model comparisons: Evaluated classical models (Logistic Regression, SVM) vs deep models (BERT, RoBERTa, etc.), comparing accuracy, F1-score, and inference time.
- Pipeline & evaluation: Incorporated cross-validation, hyperparameter tuning, and confusion matrices / ROC curves for interpretability and robustness.
A concise sentence or two about what the project does, what it’s for.
- Feature 1: core feature, novelty, or architecture.
- Feature 2: what makes it interesting or challenging.
- Feature 3: results, evaluations, or use-cases .
This project is focused on collecting, organizing, and analyzing digital collectible (NFT) data across blockchains. It enables data-driven insights into NFT holdings, transfers, and metadata trends.
- Blockchain data ingestion: Interfaces with Ethereum (and possibly other chains) via Web3, pulling transfer events, token ownership, and collection metadata.
- Metadata aggregation & cleansing: Fetches and normalizes metadata (IPFS, JSON URLs), handles missing fields, and aligns attributes across collections.
- Analytics & trend detection: Offers dashboards or scripts to analyze holder behavior, trade volumes, rarity distributions, and temporal patterns in collection activity.