All Projects
A comprehensive collection of AI, cloud, and network engineering projects spanning over 13 years of experience.
Bridgit Social Networking App & Real-Time Matching Pipeline
Sub-1s inference, 70% cloud cost reduction
Built an AI-driven, location-based, low-latency recommendation engine with sub-1s inference and 70% lower cloud cost.
Technologies
GenAI Support Automation Platform (LLMs + RAG, MLOps)
70% cost reduction, 90% faster ticket resolution
Reduced cost 70% and accelerated ticket resolution 90% by building real-time automation, CI/CD, and observability across incident workflows.
Technologies
ONAP Orchestration, Multi-Region
10× efficiency improvement, 1,000+ services scaled
Automated on-demand L2/L3 services (10× efficiency) and scaled to 1,000+ services, enabling parallel test/simulation and continuous integration.
Technologies
IZO Multi-Cloud Connect (25+ Data Centers)
Sub-2ms latency, 25+ DCs connected
Delivered sub-2ms HA interconnect supporting large-scale ML workloads and global reliability.
Technologies
LSTM Capacity Forecasting
$500K/yr savings, 92% accuracy
Built upon techniques from Kaggle's 'Predict Future Sales' competition, adapting time-series forecasting models to predict network core and edge bandwidth capacity. Achieved 92%/90% accuracy (train/test) and saved $500K/yr by preventing over-provisioning. Standardized Spark/Python feature engineering pipelines for processing historical traffic data, seasonal patterns, and network topology features.
Technologies
Python Automation SDK (5+ Teams)
80% reduction in test case development time
Cut test case development 80%; unified job execution for CI and reproducible experiments.
Technologies
OpenStack Cloud Lab
70% utilization increase, $250K/yr savings
Increased utilization 70%, saved $250K/yr; enabled high-throughput ML R&D and simulation.
Technologies
Kubernetes/Docker Orchestration & VM Consolidation
Improved runtime efficiency and throughput
Scaled test/simulation environments; improved runtime efficiency and throughput.
Technologies
Service Capacity Forecasting Analytics (4,000+ Customers)
4,000+ customers served, bottlenecks prevented
Prevented bottlenecks with distributed analytics pipelines.
Technologies
Automated Multi-Vendor Testing (Cisco, Ciena, Adva)
80% effort reduction, 3 weeks → 2-3 days
Reduced manual effort 80%; cut qualification from 3 weeks to 2–3 days.