Currently:
AI Engineer & CTO, Autipal
Based in:
Nepal (--:--)
AI engineer. I train models and ship them as real products.I'm an AI engineer. I train and fine-tune models, build RAG systems and agents, and ship them as real products. Right now that's Autipal, helping children with autism communicate in 64 languages.
Sunab Ranjitkar
01About
Hi, I'm Sunab, an AI engineer. I train models and turn them into products people use.


I'm the CTO at Autipal, an app that helps children with autism talk, learn and feel calm, in 64 languages. Right now I'm making it agentic, so the AI does more of the work for parents.
Before that I built NoaiGPT on my own. It grew to 70,000 users and was acquired in 2025. That's where I learned to train and fine-tune AI models.
02Selected work
- 01
Autipal
2026AI communication, learning and calm for children with autism. 64 languages, four platforms, going agentic.
- AI agents
- iOS
- Android
- Web
- 02
NoaiGPT
2024AI writing, humanizer and detector. 70K+ users, Microsoft for Startups, acquired in 2025.
- GPT fine-tuning
- RoBERTa
- T5
- LLaMA
03Experience
Jun 2026 — Now
Chief Technology Officer
Autipal
Hayward, California · Full-time
Leading technology for an AI app that helps children with autism communicate, learn and feel calm across iPhone, iPad, Android and the web. Now making it agentic so the AI carries more of the load for parents.
- Technology Leadership
- AI Agents
- Mobile & Web
Jan 2024 — Jan 2025
Founder
NoaiGPT
Solo founder · Acquired 2025
Built and ran an AI writing, humanizer and detection tool alone, growing it past 70,000 users through SEO and word of mouth. Fine-tuned OpenAI GPT models for the writing engine; trained RoBERTa, T5 and LLaMA on HuggingFace for the detector; built the data, backend, app, billing and servers.
- LLMs
- NLP
- Fine-tuning
- Azure
Aug 2024 — Nov 2024
Back End Developer
Nest Nepal
Lalitpur · Internship
Built and fixed server-side features at a web hosting company, and learned that “it works on my machine” is not an acceptable answer in production.
- TypeScript
- TypeORM
- Node.js
Mar 2020 — Nov 2022
QBasic / C++ Trainer
Chelsea International Academy
Kathmandu · Part-time
Taught C++ to high-school students two days a week: variables, loops, functions, and a lot of fixing missing semicolons together.
- Teaching
- C++
Education — Islington College, Kathmandu
04Tech stack
01 · the core
AI & ML
From training data to a fine-tuned model to an agent in production. These are the skills I used to grow NoaiGPT to 70,000 users and to make Autipal agentic.
38 skills · 6 areas
Fine-tuning & training
Taking a base model and making it mine
- OpenAI GPT fine-tuning
- LoRA · QLoRA · PEFT
- Full fine-tuning
- HuggingFace Transformers & Trainer
- PyTorch
- Training-data curation & labelling
- Hyperparameter tuning
LLMs & models
Picking the right model for the job
- OpenAI GPT
- Claude
- Gemini
- LLaMA
- Mistral
- RoBERTa · BERT
- T5
RAG & retrieval
Grounding answers in real data
- Embeddings
- Vector DBs: pgvector · Pinecone · Chroma
- Chunking strategies
- Hybrid search (BM25 + vectors)
- Re-ranking
- LangChain · LlamaIndex
Agents & tooling
Letting the AI do the work
- Agentic workflows
- Tool & function calling
- MCP servers
- Structured outputs
- Prompt engineering
- Multi-step planning
NLP & evaluation
Knowing when the model is right
- AI-text detection
- Text classification
- Toxicity detection
- Evals & benchmarking
- Tokenisation & long-text splitting
- Multilingual NLP (64 languages)
Deploy & MLOps
Shipping models to real users
- Azure AI · Azure ML
- GPU training jobs
- Model serving & inference APIs
- Latency & cost optimisation
- Jupyter · scikit-learn
- pandas · NumPy
02
Backend
- TypeScript
- Node.js
- TypeORM
- Express
- Python
- REST & WebSocket
- Strapi
03
Product
- Next.js
- React
- Tailwind CSS
- iOS · iPadOS
- Android
- C#
04
Ship & grow
- Microsoft Azure
- Servers & hosting
- Billing
- SEO
- Git & GitHub
Languages I work in
share of the code I write- Python 70%
- TypeScript 15%
- JavaScript 10%
- C# 5%
05Contact
© 2026 Sunab RanjitkarBack to top ↑