Hi there, I'm Vedik Agarwal 👋
GitHub • LinkedIn • Email • Featured Project
🎓 Master’s in Computer Science (Machine Learning Track) at Columbia University
🧠 ML & Systems Engineer working on LLM inference, federated learning, and ML infrastructure
⚡ Interested in memory-efficient attention, distributed optimization, and production ML systems
📄 IEEE-published researcher in federated intrusion detection
- Long-context & memory-efficient LLM inference
- Federated learning under non-IID distributions
- Distributed ML systems & microservices
- Privacy-preserving ML (Healthcare & IoT)
- High-performance model serving
Languages
ML / DL
Systems & Infra
Databases / Cloud
🚀 Featured Projects & Publications
🔗 https://github.com/vedik2002/PagedFlexAttention
- Integrated paged KV caching with PyTorch FlexAttention
- Supports long-context (4K tokens) & high-concurrency decoding
- <1% GPU memory overhead with accuracy parity on MMLU & TruthfulQA
🔗 https://ieeexplore.ieee.org/document/10857281
- Physics-based hyperparameter optimized federated MLP
- Robust under non-IID distributions
- Achieved 98% packet classification accuracy
🏭 Industry Experience
Software Developer Intern — Trademarkia
- Distributed Golang microservices scaling to 50K+ requests/day
- Low-latency BERT inference service (~200ms)
- Vector search backend using Qdrant + AWS (98% accuracy)
Software Developer Intern — Hi Rapid Lab
- AI healthcare platforms for 10K+ rural patients
- DPDP-compliant secure medical data systems
Machine Learning Intern — Innova Point Infotech
- Siamese-network-based quality inspection (PyTorch, ResNet)
- Reduced defect detection time by 45%
📊 GitHub Analytics
- LinkedIn: https://linkedin.com/in/vedik-agarwal
- Email: va2565@columbia.edu
Interested in ML systems, LLM inference, or research collaboration?
Feel free to reach out.