Skip to content
View poojakira's full-sized avatar

Block or report poojakira

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
poojakira/README.md

Capsule Banner

LinkedInPortfolioEmail

Typing SVG


⬣ PROFESSIONAL OVERVIEW

I am a Security-focused ML Engineer and IEEE-published researcher specializing in fault-tolerant GPU training and telemetry systems for aerospace and satellite workloads.

  • Target Roles: ML Engineer, ML Infrastructure / MLOps, Security-focused Cloud Engineer.
  • Core Expertise: GPU reliability, mission-control style ML infrastructure, and secure deployment pipelines.
  • Technical Stack: PyTorch (profiling & optimization), Python, Docker, Kubernetes, AWS, and Cloud Security.

🚀 Engineering Highlights

  • GPU Memory Defragmenter: Built a Transformer-driven predictor to reduce OOM failures and improve utilization on RTX-class hardware.
  • PulseNet: Developed a secure predictive-maintenance pipeline using NASA C-MAPSS data with integrated encryption and audit logging.
  • CommandX: Designed mission-control stacks for satellite telemetry, autonomous GNC, and real-time health monitoring.

🎓 Academic & Research

  • Education: M.S. in Information Technology Security, Arizona State University.
  • Fellowship: Technology Innovation Fellow with Honeywell Aerospace Labs.
  • Publications: IEEE INDICON (Adaptive EV charging) and IEEE (GPU memory optimization).

⬣ CORE TECHNICAL INFRASTRUCTURE

Area Focus
Machine Intelligence PyTorch (custom memory profiling), PINNs, Reinforcement Learning, Time-Series Modeling.
High-Performance Compute Python, SQL, C++; GPU-aware training loops; Latency-sensitive model serving.
Cloud & Security AWS (VPC, IAM, CloudWatch), Docker, Kubernetes, Applied Cryptography, Zero-Trust.
Tooling & Observability Git, GitHub Actions, Streamlit, TLE Tooling, Security Auditing.


⬣ MISSION-CRITICAL PROJECTS

🧠 Predictive GPU Memory Defragmenter

Transformer-based predictor that models GPU memory fragmentation and proactively defragments training workloads.

  • Impact: Reduced OOM failures on NVIDIA RTX 4060 testbeds by 43.6% and improved GPU utilization by 6.4%.
  • Tech: Custom PyTorch memory profiling, time-series modeling, targeted compaction.

🛰️ Apex-Aegis Tactical Suite

Physics-Informed Neural Network (PINN) engine simulating multi-physics trajectories under strict safety bounds.

  • Impact: Sub-millisecond latency for trajectory states with risk-aware constraints.
  • Tech: PINNs, RK4 Integrator, ISA Atmosphere modeling.

📡 CommandX — Satellite Mission Control

End-to-end stack unifying orbital physics, autonomous GNC, and telemetry visualization.

  • Impact: Real-time streaming of orbital elements and anomaly flags for mission-critical decision making.
  • Tech: Python, SGPL/TLE Tooling, Real-time Dashboards.

Repositories: PulseNet · CommandX · GPU Defragmenter


⬣ ACTIVITY & TELEMETRY

Pinned Loading

  1. PulseNet PulseNet Public

    AI Diagnostics: Predicts jet engine failures using NASA C-MAPSS data and Isolation Forest AI. Encrypted Pipeline: Secures telemetry via AES-256; data decrypts only in-memory. Blockchain Audit: Uses…

    Python

  2. CommandX CommandX Public

    A flight-ready satellite mission control system built on real orbital physics, autonomous GNC, and AI-driven trajectory optimization. CommandX brings together the tools mission operators need — fro…

    Python

  3. orbit-Q orbit-Q Public

    OrbitIQ Mission Control: Autonomous Satellite Health & Telemetry MLOps.

    Python

  4. pooja-portfolio pooja-portfolio Public

    Modern ML Engineer portfolio website with animated skills showcase and project gallery

    JavaScript

  5. poojakira poojakira Public

    IEEE-published researcher building scalable AI at the intersection of space & ML

  6. Predictive-GPU-Memory-Defragmenter Predictive-GPU-Memory-Defragmenter Public

    A production-grade Transformer-driven system that predicts GPU memory fragmentation before it causes OOM errors — and proactively defragments in real time

    Python