Blog freshness: Research notes liveLatest update: May 2026Telemetry mode: Public-safe live stripAI tools: Self-hosted demos live
Skip to main content

Neural Experience Layer

Impact Signals

Evidence-driven outcomes across production ML delivery, infrastructure engineering, and applied AI research — measured, not estimated.

Career Impact

Highlights by Role

Specific, verifiable outcomes from each position and project — numbers pulled directly from deliverables.

Data Scientist — GenAI & ML

Enigma Technologies · Jun 2025 – Present

  • 30%+ analyst review time saved via multi-stage RAG pipeline (OCR → retrieval → extraction → routing)
  • 20% scoring precision gain — regression, classification & clustering on 10 M+ customer records
  • LLM cost optimization via model routing (small vs premium) and semantic caching at equivalent quality
  • Cloud-native ML platform on AWS: Docker, Kubernetes, GitHub Actions CI/CD, MLflow, drift detection

Machine Learning Engineer

Cognizant · Jun 2019 – Jul 2023

  • 35% reduction in manual legal & compliance review — Contract Intelligence NLP pipeline across 10 M+ records
  • NLP latency cut from 450 ms to under 90 ms through batch inference and model optimization
  • 25% monthly AWS cost reduction via Lambda & SageMaker autoscaling; p99 latency down 40%
  • 30% faster ML release cycles — Jenkins CI/CD with automated retraining, versioning & drift alerting
  • PySpark fault-tolerant pipeline processing 1 M+ text records/day at consistent throughput

AI/ML Programming Intern

R/SEEK — UMBC · Jan 2025 – Jun 2025

  • 95% YOLOv8 object detection accuracy for autonomous RC car navigation across dynamic environments
  • TensorFlow Lite deployed on ESP32 microcontrollers for real-time edge inference — zero cloud dependency
  • Aruco marker-based positioning integrated into autonomous navigation stack

Batcave — Private AI Cloud

Personal Infrastructure · 2023 – Present

  • 56-container production platform: LLMs (LLaMA, Mistral, Phi, DeepSeek via Ollama), RAG pipelines, media, monitoring
  • 177 GB multi-database data layer — PostgreSQL, MariaDB, Redis, SQLite, Meilisearch
  • 99.9% uptime; Authentik SSO, Tailscale VPN, Netdata + Telegram real-time alerting
  • Immich ML photo platform (401 GB, face recognition + object detection), Paperless-NGX OCR

Wildfire Detection CNN

Academic Project — UMBC · 2024

  • 91% accuracy and 88% F1-score — AllCNN on Sentinel-2 multispectral satellite imagery
  • 2,400+ image patches processed with NDVI and NBR spectral indices for burn severity quantification
  • Ablation studies and k-fold cross-validation for rigorous model evaluation

Infrastructure Security Audit

Personal Project · 2026

  • Automated security scanner across 56-container production stack using Python & Docker SDK
  • 27 hardcoded credential exposures identified and remediated across services
  • CWE/CVE-aligned reporting with severity scoring and automated remediation recommendations

Good Next Move