Nishan Kumar Rai / AI Engineer

AI Engineer at HiTech Solutions and Services, Kathmandu

I build AI
that ships.

AI engineer with a full-stack and DevOps background. I build RAG systems and AI agents, the app around them and the pipeline that deploys them, bringing AI into business software used by SMEs, retailers and accountants across Nepal.

See my projectsEmail me

Drag to rotate, hover a node, or pick a group

What I build

Usually all three together: the AI, the app around it and the pipeline that ships it.

Simulated trace

Example run, not live data
loop until donePlanToolObserveDone
Try the real one
  1. Pick an example task and press Run trace.

How I work

  1. Understand

    Sit with the people who’ll use it and agree on what a good answer looks like.

  2. Prototype

    A working slice in days, on real documents or data, not a slide.

  3. Evaluate

    Test questions with known answers, check retrieval and the failure cases before anyone relies on it.

  4. Ship

    Docker, CI/CD and monitoring, so it keeps working long after the demo.

Projects

A working RAG app, plus the extraction, chat and fundamentals work behind it.

RAG application

Upload a PDF and chat with it. Every answer comes from the document, with the pages it came from.

  1. PDF
  2. Chunks
  3. Embeddings
  4. ChromaDB
  5. Llama 3.1
  6. Answer
  • Python
  • LangChain
  • ChromaDB
  • Hugging Face
  • Llama 3.1
  • Groq
  • Streamlit
Try it liveRead the case study

LLM extraction

A chatbot that reads resumes and pulls out the key details.

  • Python
  • LLM
Read the case study

LLM extraction

Extracts structured information from book content.

  • Python
  • LLM
Read the case study

Conversational AI

A conversational assistant written in Python.

  • Python
Read the case study

Notebook

A notebook of generative AI experiments.

  • Python
  • Jupyter
  • LLMs
Read the case study

Computer science

Classic data structures and algorithms implemented from scratch in Java.

  • Java
  • DSA
Read the case study

Ask my AI anything about me

A tiny retrieval system running entirely in your browser: no API and no server. It turns your question into a TF-IDF vector, finds the closest passages in my notes by cosine similarity and answers from them. Same idea as my RAG projects, minus the LLM.

Hi! I’m a small retrieval bot trained on notes about Nishan. Ask me about his work, skills, projects or how to reach him.

Under the hood: TF-IDF + cosine similarity

  1. Tokenise
  2. Vectorise
  3. Retrieve
  4. Answer

Query tokens, after stop words are removed

Ask something to see it tokenised.

Knowledge map: each dot is a passage, placed with PCA

Current roleTech stackAI skillsAI agentsFull-stack skillsDevOps skillsRAG AI Document AssistantLive demoOther AI projectsData structures & algorithmsProgramming languagesContactLocationGitHub activityBackgroundHow this site was built

Top passages

  1. The three closest passages will appear here with their scores.

Where I've been

Full-stack first, then DevOps, now AI.

Each stage adds a layer. None of them went away.

  1. Foundations

    Full-stack developer

    Built web apps end to end with the MERN stack and Python Django, and worked through data structures and algorithms in Java.

    • React
    • Node.js
    • Express
    • MongoDB
    • Django
    • PostgreSQL
    • REST APIs
    • Java DSA
  2. Level up

    DevOps

    Learned to ship what I built: containers, CI/CD pipelines, reverse proxies and Linux servers.

    • Docker
    • CI/CD
    • Nginx
    • Linux
    • Git
  3. Before HiTech

    Generative AI, RAG & agents

    Moved into LLMs: retrieval-augmented generation, embeddings, vector databases and agents that call tools.

    • LLMs
    • LangChain
    • ChromaDB
    • Hugging Face
    • Llama 3.1
    • Streamlit
    • Tool calling
  4. 2026 – present

    AI Engineer at HiTech

    At HiTech Solutions and Services in Kathmandu, bringing AI into business software used by SMEs, retailers, restaurants and accountants across Nepal.

    • RAG
    • AI agents
    • LLMs
    • Python

Let's build
something smart.

Got a pile of documents that should answer questions, an agent idea, or something you'd like to build together? Tell me about it.

in Kathmandu

Send a message

What's it about?

0 / 1000

Opens your email app with everything filled in.