Looking for a job

Aman User

@amanuser31

Message

I build AI-powered tools, fine-tuned LLM pipelines, and scalable full-stack systems.

India

Message

What I'm looking for

I'm looking to build practical AI/ML products, contribute to open-source work, and solve challenging problems across LLMs, machine learning, and full-stack systems.

At Atomic Arts, I build production AI tooling, including a PyTorch LoRA fine-tuning pipeline that reduced model turnaround time by 28% and training time by 30%.

I also developed FastAPI client-server systems and ComfyUI workflow abstractions that cut artist onboarding time by 75% and reduced file-operation latency across Windows and Linux.

Previously, I shipped AI educational tools for 3M+ daily learners at Code for GovTech, fine-tuned Qwen and Llama on AWS, and mentored four international teams at Sugar Labs on responsible AI/ML practices.

Experience

Work history, roles, and key accomplishments

Current

Software Developer

Current

Atomic Arts

Jul 2026 - Present (2 months)

Developed a proprietary pipeline tool integrating PyTorch-based LoRA fine-tuning with automated dataset ingestion and GPU job routing, reducing model turnaround time by 28% and training time by 30%. Built a Python wrapper layer using the ComfyUI API and a distributed FastAPI client-server system with load balancing and cross-platform file hardlinking.

PyTorch LoRAs fastAPI REST APIs ComfyUI Electron Python GPU Load Balancing Windows Linux

Google Summer of Code 2026 Mentor

Sugar Labs

May 2026 - Aug 2026 (3 months)

Evaluated contributor proposals against technical and ethical AI criteria and guided 4 teams on responsible AI/ML practices. Led weekly syncs, unblocked architectural decisions, and co-authored onboarding documentation.

AI Mentorship Technical Evaluation Documentation Communication Skills

AI Software Developer

Code for GovTech

Jun 2025 - Sep 2025 (3 months)

Built AI-powered educational tools for 3M+ daily learners across 40+ countries, shipping an AI Story Builder feature. Fine-tuned and deployed LLMs on AWS SageMaker and EC2, optimizing inference response time from 4.0s to 0.5s through model quantization.