Venture + Data

• Designed and implemented agentic AI workflows using LLMs to automate key venture capital operations, improving decision-making speed and consistency
• Developed data pipelines for deal sourcing and portfolio analysis using Python and SQL
• Supported VC activities including market research, startup screening, and due diligence through data-driven insights

Full-stack development on AI-powered recruitment platform, executing complete SDLC from design to deployment

• Integrated new software components into the impress.ai platform with comprehensive documentation
• Troubleshooted and debugged existing systems while implementing performance optimizations

Undergraduate research on Explainable AI for Large Language Models under Prof Shen Zhiqi

• Evaluated 3+ explanation methods (SHAP, LIME, attention visualization) for LLM interpretability
• Developed interactive frontend dashboard for real-world user testing of explanation interfaces
• Contributing to research paper on bridging AI interpretability and usability for non-technical users

Built CNN-based classification system for diabetic foot thermal imaging diagnostic workflows

• Preprocessed 100+ thermal images with normalization and contrast enhancement techniques
• Developed custom CNN architecture achieving high precision/recall for medical screening automation

Financial advisory and portfolio management at mutual fund distribution firm

• Analyzed client financial goals to recommend tailored mutual fund portfolios using Ngen and Excel
• Created 5+ client presentations on portfolio diversification strategies and market trends

IoT device identification research using network traffic analysis and machine learning

• Improved device identification accuracy by 15% using hybrid rule-based and ML approaches
• Processed packet captures (PCAPs) to extract features like MAC patterns and protocol distributions
• Developed meta-classifier combining multiple identification methods for robust real-time analysis

VR training simulation for field hockey goalkeepers using Unity and Meta Quest 3

• Built immersive gameplay with randomized ball trajectories and full-body avatar tracking
• Optimized performance for 90+ FPS rendering with multiple difficulty modes and feedback systems

Client project development and technology innovation collaboration

• Enhanced 2+ client project outcomes through strategic planning and technical implementation
• Contributed to high-stakes decision-making processes under departmental leadership guidance