Experience
Building production AI systems and research prototypes
Co-Founder
HyperSentry
ML Systems Engineer (Co-Founder)
- Built a security copilot using OPA/Rego policy checks + RAG over security knowledge bases, auto-creating remediation PRs in GitHub/AWS
- Held false positives under 3% while auto-merging 30%+ of fixes; processed 50+ PRs/week with 24h remediation SLA
- Optimized retrieval + reasoning pipeline (embedding cache, batched inference, parallel I/O) to keep E2E <800ms under typical load
- Integrated Slack and CLI interfaces for natural language queries, auto-generated CloudTrail/Steampipe queries for blast radius analysis
PythonRAGOPA/RegoGitHub APIAWSCloudTrailSteampipeSlack
Full-time
Zscaler
Machine Learning Engineer
🤖ZDX Copilot
- Led RAG + LLM Guardrails development: fine-tuned embeddings to reduce hallucination rates by 18%, increasing ISP incident detection accuracy by 30%
- Implemented contrastive loss objective for network anomaly detection, improving few-shot learning adaptability by 12%
- Employed runtime and memory profiling to reduce peak memory from 220GB to 52GB, decreasing compute costs by 6×
- Engineered hybrid Spark + Kubernetes architecture that expanded training data capacity 4×, accommodating larger customer deployments
PythonRAGBERTContrastive LearningSparkKubernetesLLM Guardrails
🔐ZPA Intelligent Policy
- Patented Community Detection: utilized sequential user-app access data to construct network graphs via Louvain algorithm, achieving 80% adoption of top 100 policy recommendations
- Enhanced policy recommendation engine through error analysis, EDA, and ML (K-means, HDBSCAN, Random Forests), resulting in 23% reduction in policy violations
- Built graph-based policy recommender analyzing access patterns to automate security policy configuration
PythonGraph MLLouvain ClusteringK-meansHDBSCANRandom Forests
Looking for a detailed resume?
Download Resume