AI/ML Engineer • Agentic AI advocate • ML Infrastructure • LLM Agents • Production ML

Creating intuitive AI systems designed for everyone

I build autonomous AI systems capable of independently performing machine learning engineering tasks — from raw data ingestion to production deployment.

My work focuses on replacing manual ML workflows with intelligent, self-optimizing systems designed for real-world deployment.

Core focus areas:

• Autonomous ML agents
• LLM-powered reasoning systems
• Production-grade ML pipelines
• Efficient AI systems optimized for constrained hardware
• Modular, explainable AI architectures

These systems are designed as real engineering infrastructure, not experimental prototypes.

Python • scikit-learn • PyTorch • Optuna • NumPy • Pandas

ML orchestration • Autonomous pipelines • LLM systems • Strategy memory systems

Flask • Joblib • CLI systems • Docker-ready architectures • Modular pipeline design

Production ML systems
Autonomous pipeline design
Model automation and orchestration
Compute-efficient AI engineering

Autonomous Machine Learning Engineer

Repository:
https://github.com/Mihawk1891/TuneLab

TuneLab is a fully autonomous machine learning agent that builds production-ready ML models directly from raw CSV datasets without human intervention.

Core capabilities:

• Automatic dataset analysis and target detection
• Autonomous feature engineering and preprocessing
• Multi-model training and evaluation
• Hyperparameter optimization using Optuna
• Deployment-ready artifact generation
• Automated report and documentation generation
• Strategy memory system using dataset fingerprinting
• Incremental improvement through past learning reuse

Autonomous workflow architecture:

Raw Dataset
    ↓
Dataset Fingerprinting
    ↓
Data Understanding
    ↓
Feature Engineering
    ↓
Multi-Model Training
    ↓
Hyperparameter Optimization
    ↓
Model Artifact Generation
    ↓
Strategy Memory Storage

Engineering highlights:

• Fully autonomous ML pipeline
• CPU-optimized, no GPU required
• Production deployment ready
• Strategy reuse across datasets
• Modular and extensible architecture
• ~1,500 lines production-grade code

This system functions as an autonomous ML engineer capable of independently producing deployable machine learning systems.

Autonomous GAN-Based Image Restoration System

Repository:
https://github.com/Mihawk1891/unpotatofy

Unpotatofy is a production-ready image enhancement pipeline using lightweight GAN architectures optimized for low-memory environments.

Core capabilities:

• Intelligent image quality analysis using classical CV metrics
• Conditional pipeline execution based on detected defects
• Automatic enhancement pipeline construction
• Fully modular enhancement architecture
• Fully offline operation

Integrated enhancement models:

• NAFNet — motion blur restoration
• Real-ESRGAN — super-resolution enhancement
• DeOldify — grayscale image colorization

Pipeline architecture:

Input Image
    ↓
Quality Analysis
    ↓
Conditional Pipeline Builder
    ↓
Sequential Enhancement Execution
    ↓
Enhanced Output

Engineering highlights:

• Runs on 4GB GPU or CPU
• Sequential model loading to minimize memory usage
• Predictable and stable memory footprint
• Modular enhancement architecture
• Production-grade error handling and logging
• Explainable enhancement decision pipeline
• Fully offline execution capability

This system demonstrates deployment of efficient GAN-based restoration pipelines under strict hardware constraints.

LLM-Powered Academic Intelligence System

Repository:
https://github.com/Mihawk1891/SaarAI

SaarAI is an intelligent system that uses large language models to automate academic analysis and structured report generation.

Capabilities:

• Automated academic data processing
• LLM-based reasoning pipeline
• Structured report generation
• Modular AI pipeline architecture
• Autonomous insight generation

Engineering focus:

LLM pipeline design
Autonomous reasoning workflows
Structured AI output systems

All systems are designed using the following engineering principles:

• Autonomous operation over manual control
• Production-ready system architecture
• Efficient hardware utilization
• Modular and extensible system design
• Explainable and deterministic behavior
• Real-world deployment readiness

Goal: build AI systems that function as independent engineering entities.

Actively building systems capable of operating as:

• Autonomous machine learning engineers
• Self-optimizing ML pipelines
• LLM-powered reasoning agents
• Production AI infrastructure components

Focus is on autonomous intelligence capable of performing real engineering tasks.

Email: pranavbansode2604@gmail.com
LinkedIn: https://www.linkedin.com/in/pranav-bansode-281793229/
GitHub: https://github.com/Mihawk1891