Open to opportunities

Suraj Sah Kanu

@surajsahkanu

AI/ML Engineer focused on building production-ready AI systems, RAG applications, LLM agents, and machine learning solutions.

Nepal

Message

What I'm looking for

I'm looking to build scalable, production-ready AI systems that bridge applied research and deployment, particularly across LLMs, RAG, multi-agent workflows, computer vision, and MLOps.

I've delivered end-to-end AI solutions as a Freelance AI/ML Engineer, spanning machine learning, deep learning, computer vision, LLMs, RAG, and multi-agent AI. I take projects from architecture and evaluation through deployment, containerization, and documentation.

I fine-tuned Qwen3-4B-Instruct with QLoRA, PEFT, Unsloth, and TRL, producing reproducible training pipelines and LoRA, FP16, and GGUF model artifacts. I also built production-ready conversational RAG backends with FastAPI, Qdrant, LangChain, Redis, NVIDIA NIM, and Docker.

At NepAI Applied Mathematics and Informatics Institute for Research (NAAMII), I built medical image embedding pipelines for 3D CT/MRI scans and fine-tuning strategies for automated esophagitis grading that achieved 92%+ validation accuracy. My preprocessing automation reduced image processing overhead by 65%.

I've also developed CrewAI and LangGraph systems for hotel booking, workflow automation, and compliance-document intelligence, with traceable extraction and reconciliation workflows. I enjoy turning applied AI research into scalable, reproducible systems.

Experience

Work history, roles, and key accomplishments

Current

Freelance AI/ML Engineer

Current

Independent AI Engineering & Consulting

Jan 2024 - Present (2 years 7 months)

Delivered end-to-end AI solutions for independent projects and client-oriented use cases across Machine Learning, Deep Learning, Computer Vision, LLMs, RAG, and Multi-Agent AI, independently handling architecture, development, evaluation, deployment, and documentation.

Machine Learning Research Intern

NAAMII

Aug 2025 - Apr 2026 (8 months)

Built a 3D medical imaging (CT/MRI) embedding pipeline with MedicalNet ResNet, fine-tuned models for esophagitis grading (92%+ accuracy via k-fold CV), engineered a 65%-faster preprocessing pipeline, and built PCA/outlier-detection tools for embedding QA and reproducible research.

Python PyTorch OpenCV Deep Learning Foundation Models Feature Extraction Medical AI Machine Learning