Apple MLX ↗
• Contributor adding CUDA support to MLX, working on FFT CUDA Backend.

• Modified mx.sort() behavior in CPU and Metal Backends to match NaN Sorting behavior based on IEEE-754.

• Added QoL Improvements such as #2689 (Einsum Error Msg Improvement) and #2690 (Fixed Type Annotations in Stubs).

Apple Silicon RDMA Simulation Testbed

• Deployed K3s Cluster on LimaVM on a M4 Mini, using it to run AI/ML Workloads

• Enabled Soft-RoCE to simulate RDMA behavior, set up Cilium Hubble for eBPF enabled observability of networking.

Gear9 ↗

• Designed a Digital Pitwall which works with F1 23 Videogame, streaming the game data via UDP.

• Created Data Streaming Pipeline hosted on AWS for real time data during racing.

• Data stored and collected on Prometheus, with all Data Visualization on Grafana.

Guide for Preparing and Responding to Deepfake Events ↗

(Co-authored with OWASP Members)

Defined deepfake incident-response

playbooks with detection, containment, eradication, and recovery workflows for fraud, impersonation, and misinformation scenarios.

GenAI Red Teaming Guide ↗

(Co-authored with OWASP Members)

Defined deepfake incident-response

Designed an adversarial testing blueprint (Model, Implementation, System, Runtime) integrated into CI/CD to fortify AI services against attacks. 2nd most downloaded OWASP Top 10 for LLMs White paper.

LLM & GenAI Data Security Best Practices ↗

(Co-authored with OWASP Members)

Defined defense measures including layered

encryption, SIEM/XDR monitoring, secure data flows for agentic LLMs and governance frameworks to ensure integrity and reliability of AI data pipelines.