🚀 Software Developer | AI/ML & LLM Enthusiast | Backend Engineer

Welcome to my GitHub! I’m a curious developer who enjoys going beyond simply using technology — I like understanding how things work under the hood, experimenting with new approaches, and turning ideas into practical systems.

My current focus is around AI/LLM engineering, backend systems, model serving, and automation.

• 🤖 Generative AI & Large Language Models
• 🧩 LLM Fine-Tuning & Continuous Learning
• ⚡ High-Performance Model Serving & Inference
• 🏗️ Backend Engineering & System Design
• 🔄 Distributed Systems & Background Workers
• 🐳 Docker & Containerized Architectures
• ☁️ Cloud Infrastructure & Scalable Services
• 🔓 Open Source & Developer Tools
• 🛠️ Automation and Developer Productivity

Working with modern LLM workflows involving:

• Fine-tuning with SFT, DPO, LoRA and QLoRA
• Building feedback-driven continuous learning pipelines
• Dataset preparation, evaluation, and model comparison
• RAG + fine-tuning approaches
• Model quantization and efficient inference
• Production deployment of open-weight models

Experimenting with high-performance inference stacks such as:

• vLLM
• GPU-optimized inference
• KV cache optimization
• Prefix caching
• Chunked prefill
• Speculative decoding
• FP8 / quantized inference
• Benchmarking throughput, latency, TTFT and concurrency

I've been experimenting with models such as Gemma and Muse-Glimmer, including fine-tuning and production-style serving workflows.

I’m also working on improving backend architecture through:

• FastAPI
• Service-oriented architectures
• Independent worker services
• Docker-based deployments
• Designing systems that are easier to deploy, scale, monitor, and maintain

Python · JavaScript · SQL

PyTorch · Transformers · PEFT · LoRA · QLoRA · SFT · DPO · RAG

vLLM · Ollama · Hugging Face · Quantization · GPU Inference

FastAPI · REST APIs · MongoDB · Vector Databases

Docker · AWS · AWS SQS · Linux · Git · GitHub

Jupyter · VS Code · PySide6 · FFmpeg

A few areas I’m actively working on:

🧠 Feedback-Driven LLM Learning

Building workflows where model-generated responses are collected with user feedback, converted into structured datasets, and periodically used for fine-tuning and evaluation.

⚡ LLM Inference & Benchmarking

Testing different serving configurations and model optimizations to understand the trade-offs between latency, throughput, concurrency, and GPU memory usage.

🛠️ Developer Tools

Building practical utilities for media processing, subtitle workflows, automation, and developer productivity.

I’m continuously improving my understanding of:

• System design and distributed architectures
• Advanced Python and backend engineering
• Machine learning fundamentals
• Statistics and data analysis
• Data science workflows
• LLM evaluation and optimization
• Open-source development and contribution practices

I’m actively interested in contributing to Python, AI, ML, and LLM-related open-source projects.

I particularly enjoy projects where I can:

• Fix real-world bugs
• Improve developer experience
• Improve documentation
• Build useful tooling
• Learn from production-grade codebases

I believe the best way to learn engineering is to build, break, debug, improve, and contribute.

Understand the problem
        ↓
Build a simple solution
        ↓
Measure it
        ↓
Find the bottleneck
        ↓
Optimize it
        ↓
Automate it
        ↓
Ship it 🚀

I care about clean code, measurable performance, maintainable architecture, and solutions that actually work in production.

• 🐍 Python is one of my main tools for building things
• 🤖 I enjoy experimenting with LLMs more than just consuming them
• ⚡ I like benchmarking systems instead of guessing about performance
• 🧠 I enjoy solving logic and engineering problems
• 🔧 I have a habit of revisiting old code and finding ways to improve it
• 🌑 And yes... dark mode is life. 😎

📧 Email: satheesh142002@gmail.com

💼 LinkedIn: linkedin.com/in/satheesh142002

“First, solve the problem. Then, write the code.” — John Johnson

⭐ Thanks for stopping by my profile!