Hero background

    Ason Gautam

    AI ENGINEER AND FULLSTACK DEVELOPER.

    # AI & Full Stack Expertise
    
    import torch
    from transformers import AutoModel, AutoTokenizer
    from langchain.vectorstores import FAISS
    from react import {useState, useEffect}
    import next.server as nextjs
    
    class Expertise:
      def __init__(self):
        self.ai_skills = ["RAG", "LLMs", "Fine-tuning", "Prompt Engineering"]
        self.dev_stack = ["React", "Next.js", "Node.js", "TypeScript", "Python"]
        self.design_tools = ["Figma", "Photoshop", "Canva"]
        self.cloud_platforms = ["AWS", "Docker", "Kubernetes"]
    
      def build_rag_system(self, model_name: str):
        try:
          tokenizer = AutoTokenizer.from_pretrained(model_name)
          model = AutoModel.from_pretrained(model_name)
          return {"status": "Production-ready AI"}
        except Exception as e:
          return {"error": str(e)}
    
      def create_fullstack_app(self, frontend: str, backend: str):
        return "Building fullstack application"
    
      def design_ui(self, tool: str):
        if tool in self.design_tools:
          return "Creating beautiful UI"
        return "Design tool not found"
    
    # Always learning, always building
    expertise = Expertise()
    rag_result = expertise.build_rag_system("llama-2-7b")
    app = expertise.create_fullstack_app("Next.js", "Node.js")
    design = expertise.design_ui("Figma")
    print("🚀 Ready to build")
    total_skills = len(expertise.ai_skills + expertise.dev_stack)
    print("Skills: " + str(total_skills) + " technologies")

    My Expertise

    AI Engineering

    LLMs & AI Systems

    Expert in building RAG systems, developing AI agents, and fine-tuning LLMs. Strong in prompt engineering, model deployment, and creating production-ready AI solutions at scale.

    Full Stack Development

    Frontend, Backend & DevOps

    Creating complete web applications from database to UI. Proficient in React, Next.js, Node.js, Laravel, Python, and modern cloud infrastructure. Experienced in CI/CD, containerization, and DevOps practices.

    Designing

    UI/UX & Graphics

    Crafting beautiful, user-centric designs with attention to detail. Proficient in Figma, Canva, and Photoshop. Leveraging AI tools to create stunning visuals and seamless user experiences.

    
    import React, {useState, useEffect, useMemo} from "react"
    import Image from "next/image"
    import Link from "next/link"
    import {ExternalLink, Github, Code} from "lucide-react"
    
    type ProjectCategory = "ai" | "web-dev" | "all"
    type ProjectStatus = "live" | "development" | "archived"
    
    interface Project {
      id: number
      title: string
      description: string
      image: string
      category: ProjectCategory
      status: ProjectStatus
      link?: string
      github?: string
      technologies: string[]
      year: number
    }
    
    const CATEGORIES: ProjectCategory[] = ["all", "ai", "web-dev"]
    
    const projects: Project[] = [
      {
        id: 1,
        title: "Paperight.ai",
        description: "AI-powered final year project analysis",
        category: "ai",
        status: "development",
        technologies: ["Next.js", "LLMs", "RAG", "TypeScript"],
        year: 2025
      },
      {
        id: 2,
        title: "MyPusta",
        description: "Nepali family tree & vamsavali builder",
        category: "web-dev",
        status: "live",
        link: "https://mypusta.com",
        technologies: ["React", "Node.js", "PostgreSQL", "Express"],
        year: 2024
      },
      {
        id: 3,
        title: "CloudModel Nexus",
        description: "Cloud platform for LLM hosting",
        category: "ai",
        status: "live",
        technologies: ["Python", "Docker", "Kubernetes", "FastAPI"],
        year: 2024
      },
      {
        id: 4,
        title: "ARMS 360",
        description: "Deerwalk school ERP",
        category: "web-dev",
        status: "live",
        link: "https://arms360.deerwalk.edu.np/login",
        technologies: ["Next.js", "TypeScript", "Tailwind", "Framer Motion"],
        year: 2024
      },
      {
        id: 5,
        title: "Music Jam Space",
        description: "Rehearsal studio booking platform",
        category: "web-dev",
        status: "live",
        link: "https://musicjamspace.com",
        technologies: ["React", "Laravel", "MySQL"],
        year: 2023
      }
    ]
    
    function filterProjects(
      projects: Project[],
      category?: ProjectCategory
    ) {
      if (!category || category === "all") {
        return projects
      }
      return projects.filter(
        (project) => project.category === category
      )
    }
    
    function sortProjectsByYear(
      projects: Project[]
    ) {
      return [...projects].sort((a, b) => b.year - a.year)
    }
    
    function getProjectStats(
      projects: Project[]
    ) {
      const live = projects.filter(p => p.status === "live").length
      const inDev = projects.filter(p => p.status === "development").length
      return { live, inDev, total: projects.length }
    }
    
    export default function Work() {
      const [selectedCategory, setCategory] = useState<ProjectCategory>("all")
      const [sortBy, setSortBy] = useState<"year" | "name">("year")
    
      const filteredProjects = useMemo(
        () => {
          let result = filterProjects(projects, selectedCategory)
          if (sortBy === "year") {
            result = sortProjectsByYear(result)
          }
          return result
        },
        [selectedCategory, sortBy]
      )
    
      const stats = useMemo(
        () => getProjectStats(filteredProjects),
        [filteredProjects]
      )
    
      useEffect(() => {
        console.log("Projects loaded:", filteredProjects.length)
      }, [filteredProjects])
    
      return (
        <section className="work-section">
          <div className="header">
            <h2>My Work</h2>
            <div className="stats">
              <span>{stats.live} Live</span>
              <span>{stats.inDev} In Dev</span>
            </div>
          </div>
          <div className="filters">
            {CATEGORIES.map(cat => (
              <button
                key={cat}
                onClick={() => setCategory(cat)}
                className={selectedCategory === cat ? "active" : ""}
              >
                {cat}
              </button>
            ))
          </div>
          <div className="projects-grid">
            {filteredProjects.map((project) => (
              <ProjectCard
                key={project.id}
                title={project.title}
                description={project.description}
                status={project.status}
                technologies={project.technologies}
                link={project.link}
              />
            ))
          </div>
        </section>
      )
    }

    My
    Work

    Delivered scalable solutions for domestic and international clients, from travel platforms to telemedicine systems. Specialized in React, PWA, and modern web technologies, covering both traditional web applications and AI-powered solutions.

    IN DEVELOPMENT

    PropertyFlow

    Currently developing property management software for a client in the real estate sector. The solution will automate rent collection, tenant management, payment tracking, and financial reporting, helping landlords and commercial building owners efficiently manage multiple properties and monthly income from tenants.

    Paperight.ai

    Paperight.ai

    AI-powered final year project analysis system for BScCSIT students under Tribhuvan University guidelines. Features an intuitive drag-and-drop report upload interface, archive management system, and comprehensive AI analysis that provides detailed insights, gap analysis, and actionable suggestions. Built with custom AI agents using multiple LLMs. Currently in development — launching December 2025.

    MyPusta

    MyPusta

    MyPusta helps Nepali families build and preserve their family tree — so every ancestor's name, village and story lives on. Built specifically for Nepali lineage with Devanagari support, gotra and ancestral village (गाउँ) fields, interactive tree visualization, and private collaborative editing so relatives anywhere can add the names and memories only they remember. Now in beta at mypusta.com, with 100+ families already mapping their पुस्ता across generations.

    Music Jam Space

    Music Jam Space

    Built a full booking platform for Music Jam Space — a rising Nepali music community hub in Kathmandu. Musicians and bands can rent rehearsal rooms, check schedules, and book practice sessions online. The system streamlines room availability, hourly bookings, and studio services so artists can focus on creating. Now live at musicjamspace.com, helping the local music scene practice, record, and grow with less friction.

    ARMS 360 — Deerwalk School ERP

    ARMS 360 — Deerwalk School ERP

    Enterprise school ERP for Deerwalk Education Group, live across 4+ schools. ARMS 360 centralizes academics and operations with automatic marksheet generation, attendance tracking, bus tracking, student and staff management, and role-based access for administrators, teachers, and parents. Built as a production-grade platform that streamlines day-to-day school workflows from classroom records to campus logistics.

    CloudModel Nexus

    CloudModel Nexus

    Enterprise-grade cloud platform for hosting and deploying fine-tuned large language models. Successfully deployed and managed multiple state-of-the-art LLMs including DeepSeek, ChatGPT, LLaMA, and Gemini models in scalable cloud infrastructure. Built a robust orchestration system enabling seamless model deployment, versioning, and inference management. Provides high-performance API endpoints, load balancing, and monitoring capabilities for production-ready AI model hosting. Demonstrates expertise in cloud architecture, containerization, and managing complex ML workloads at scale.

    ✨ Get access to the cloud model or get an API key at cheap rate • Contact Me

    Portfolio-as-a-Service Platform

    Portfolio-as-a-Service Platform

    Developed a comprehensive portfolio solution for clients seeking modern, professional online presence. Delivered a platform featuring modern, clean designs with customizable navigation, hero sections, and personal branding. Successfully enabled businesses and individuals to create professional portfolios without expensive custom development. Example shown: Sujata Shrestha's portfolio with elegant UI, multi-language support, and seamless content management. The platform includes custom domain integration and fully customizable branding through an intuitive interface, providing clients with a complete portfolio management solution.

    Career

    Professional Experience

    Deerwalk Institute of Technology

    AI Engineer & Team Lead

    R&D · Deerwalk Institute of Technology

    2025 — Present
    Kathmandu, Nepaldeerwalk.edu.np

    Lead a 12-person R&D team delivering AI and full-stack products. Own model fine-tuning, prompt systems, and RAG pipelines for production use cases, while shipping end-to-end software and mentoring engineers on modern AI practices.

    AI Fine-tuningPrompt EngineeringRAGFull StackLLMsPython

    Fullstack Developer

    Freelance

    2024 — Present
    Remote · Global

    Partner with startups and businesses worldwide to design and ship production web apps. Deliver AI-integrated products, modern UI systems, and scalable backends — from first prototype through launch and iteration.

    ReactNext.jsNode.jsTypeScriptAI IntegrationUI/UX Design
    Aquiden Software

    Fullstack Developer

    Aquiden Software

    2024 — 2025
    Kathmandu, Nepalaquiden.com

    Built full-stack applications across React frontends and Node.js / Laravel backends. Owned API design, SQL data models, and responsive interfaces for client products with reliable end-to-end delivery.

    ReactNode.jsExpressSQLLaravelFull Stack

    Research
    Papers

    How Chunk Size and Embedding Model Choice Affect Retrieval-Augmented Generation Quality and Cost: A Controlled Comparison on Consumer Hardware

    A controlled study of Retrieval-Augmented Generation (RAG) pipelines, comparing three chunk sizes (128, 256, 512 words) across three compact embedding models (BGE-small, E5-small, MiniLM). Measures indexing cost, retrieval quality (Recall@5), and end-to-end answer quality (F1, Exact Match) on a fixed SQuAD 1.1 benchmark with a local Llama 3.2 3B model, showing that larger chunks and retrieval-tuned embeddings tend to outperform common assumptions.

    RAGchunk sizeembeddingsdense retrievalSQuAD
    AuthorAson Gautam
    AffiliationIndependent Researcher
    LicenseCC BY 4.0
    Version1.0
    PublisherZenodo

    Available for select
    freelance opportunities

    Have an exciting project you need help with? Send me an email or contact me via instant message!

    Ason helped Music Jam Space take a real leap. With the online booking system he built, our community is growing fast and we're now getting international clients too. Practice room bookings run smoothly, musicians keep coming back, and the whole studio feels more connected than ever. He didn't just ship a website — he boosted our business.

    - Sarnil MaharjanMusic Jam Space

    Ason is AMAZING! If you’re unsure about hiring him — don’t be! He built an incredible website for PowerFlow that’s clean, fast, and exactly what we needed. We’re super impressed with his creativity, professionalism, and attention to detail!”

    - RijanPowerflow Solutions

    Ason delivered a custom ERP for our startup that transformed the way we operate. He took our scattered workflows, organized them into a single intuitive platform, and iterated quickly on feedback. The result is a reliable system our whole team depends on every day.

    - Saujanya Sharma