About Services Process Work Contact
ML • Computer Vision • Computational Physics

Hi, I'm Deepak Pandey

I build practical ML systems: custom computer vision, model training, analytics, and research-grade simulations.

📍 Darwin, Australia • Remote freelance • Research collaboration

Available for ML / CV projects • Training • Deployment • Physics simulations

About

I work at the intersection of machine learning, computer vision, and computational physics. My focus is building systems that run in real conditions: real-time detection/tracking, model training pipelines, data analytics, and scientific simulations.

If you need a model that is not “demo-only” but is designed for robustness, speed, and deployment, I can help you go from idea → dataset → training → evaluation → production.

Production mindset

Latency/FPS, failure modes, monitoring, and real workflows—not just accuracy numbers.

Real-time CV

End-to-end delivery

Data pipelines, training, metrics, packaging, and deployment to Windows/Linux.

From data to deploy

Scientific computing

Hands-on with simulations and HPC-style workflows (GROMACS, WIEN2k, Quantum ESPRESSO).

Research collaboration

Freelance Services

👁️

Computer Vision Systems

Detection, tracking, counting, measurement, ROI pipelines, multi-camera setups, and real-time inference optimization for practical environments.

🧠

Custom ML Model Development

Custom architectures and training workflows, dataset planning, evaluation metrics, ablation testing, and robustness improvements.

🏋️

Model Training & Fine-tuning

Transfer learning, domain adaptation, hard-sample mining, labeling strategy, and reproducible experiments with clear reporting.

📊

Data Analysis & Dashboards

Python analytics, SQL/Excel automation, dashboards and reporting for operations, R&D, and monitoring systems.

⚙️

Deployment & Optimization

Packaging, performance profiling, GPU acceleration planning, and deployment on Windows/Linux with reliable runtime behavior.

🔬

Physics Simulations

Collaboration-ready simulation workflows: GROMACS molecular dynamics, WIEN2k (DFT), and Quantum ESPRESSO computations.

How We Work

Step 1

Scope and success metrics

We define your goal, constraints, target metrics (accuracy/FPS), and deliverables with a clear plan.

Step 2

Data and pipeline

Dataset plan, labeling rules, augmentation strategy, baseline model, and evaluation protocol.

Step 3

Training and validation

Train, test, iterate. You get a report with metrics, failure cases, and next improvements.

Step 4

Delivery and deployment

Packaged deliverables (code + model + instructions), deployment support, and performance tuning.

Work

Selected repositories and models pulled automatically from GitHub and Hugging Face.

Loading GitHub repositories…
Loading Hugging Face models…

Tools I Use

🐍
Python
🔥
PyTorch
👁️
OpenCV
🧮
NumPy/SciPy
📊
Matplotlib
🗃️
SQL / SQLite
CUDA
⚛️
Quantum / Qiskit
🧬
GROMACS
🧪
WIEN2k
🌊
Quantum ESPRESSO

FAQ

What do you need from me to start?+
A short description of the goal, sample data (if available), constraints (hardware, time), and what “success” means (accuracy/FPS/robustness).
Can you train a custom model for my specific domain?+
Yes. I can help with dataset strategy, labeling rules, baseline training, evaluation, and improvements for edge cases until performance is reliable.
Do you also help with deployment?+
Yes—packaging, reproducible installs, performance profiling, and deployment workflows for Windows/Linux are part of how I deliver projects.

Contact

📧

Email

dpkarcai@protonmail.com

💼

Engagement

Freelance • Consulting • Collaboration

⏱️

Response

Usually within 24 hours

✓ Thank you! I’ll respond within 24 hours.
✗ Something went wrong. Please email dpkarcai@protonmail.com