Part time

Knightmare Verified

AI Engineer

Kathmandu Metropolitan City, Kathmandu, Bagamati Province, Nepal 4+ yrs experience 7 skills $50/hr
Link copied!
Download CV

About

AI Engineer with nearly 4 years of experience in data science, backend development, and advanced LLM implementation.Skilled in building and automating AI-driven systems using Python, FastAPI, and LangGraph. Specialized in RAGpipelines, LLM optimization, knowledge graphs, voice agents, and research on GraphRAG architectures. Strongdocumentation advocate with mentoring and cross-functional collaboration experience.

Experience

Mid-level Team Lead and AI Engineer (Contract)

Techtrums

Aug 2026 — Present
  • Building a pregnancy platform delivering personalized nutrition diets and health recommendations by ingesting a knowledge graph aligned to Nepal and Australian health standards.
  • Built an automatic tagging system that recommends specific products based on pregnancy status / trimester.
  • Built a voice-agent platform using Gemini models for a custom voice receptionist, with MCP integration to call health APIs from the project.

AI and Automation

Logictive Solutions

Feb 2026 — Jul 2026
  • Built video generation for the launch event of new ISP Voxtra.
  • Built Voxtra's location-availability service on PHP / CodeIgniter, storing coverage for locations across Australia in the database for faster availability querying.
  • Built an internal tool to streamline the new-client process, combining website vulnerability scanning with a custom automation form for new requirements and features.
  • Built a competitor-intelligence platform alongside the marketing team to track competitor posts, social media targets, and target audiences, with automation via n8n and custom endpoints for weekly leads and insights from the Meta Ads platform.

AI Engineer (Contract)

Awasarlogic

Feb 2026 — May 2026
  • Built a video generation module that auto-generates problem-solving videos per topic / question using Python, Manim, and an open-source voice agent.
  • Built backend for the bus module, including route management, ETA estimation, and notifications for arrival delays.

Associate AI Engineer

Frost Digital Ventures Pvt. Ltd

Feb 2024 — Feb 2026
  • Architected the "Discover" module for Frost's Ecosystem platform, integrating 3+ external APIs (Apollo, Invicta, Wikipedia) - enhancing company data coverage by 60% and supporting automated report generation.
  • Developed TrackLab, a LangGraph-powered automation app that consolidated 40+ weekly status reports - reducing manual compilation time from 2 days to under 10 minutes.
  • Introduced LangGraph orchestration into the SI8 project to streamline report column management and narrative generation, resulting in top-ranked report accuracy among internal benchmarks.
  • Designed and deployed a modular RAG (Retrieval-Augmented Generation) pipeline, leveraging vector databases (Weaviate, Qdrant) and custom prompt templates - enabling context-aware company intelligence queries with 50% faster response generation.
  • Led Sphinx documentation standardization, converting 5 major project repos into searchable, versioned docs - improving cross-team onboarding and handover efficiency by 40%.

Jr. AI Engineer

InfoDevelopers Pvt. Ltd

Dec 2022 — Feb 2024
  • Automated financial data acquisition pipelines using 25+ FastAPI microservices - cutting manual ETL work by 70% and enabling real-time integration with financial dashboards.
  • Built ARIMA and Exponential Smoothing models for transaction forecasting, achieving 90%+ prediction accuracy on time-series data and supporting credit risk assessments.
  • Developed automated trading workflow logic for the company's TMS (Trading Management System), introducing smart order execution and alert features that increased trade throughput by 30%.
  • Scraped and structured large-scale job datasets using Selenium and BeautifulSoup to support internal CV parsing R&D, building a labeled corpus of 40,000+ job entries.

Education

B.E. in Computer Science

Purbanchal University (Acme Engineering College)

Jan 2018 — Jan 2023
CGPA: 3.56 / 4.0. Relevant coursework: Artificial Intelligence, Data Mining, DBMS, Probability & Statistics, Discrete Mathematics, Advanced Computer Architecture.

Projects

Live

Waste Vehicle Tracking and Management (FYP)

Arduino, GPRS/GSM, PHP, eSewa API

Designed an IoT-based waste tracking system using Arduino (GPRS/GSM), PHP, and the eSewa API, enabling smart billing and real-time route tracking.
View project
Live

News Summarization & Classification

Python, scikit-learn, NLP, TF-IDF

Trained multi-language news classifiers using SVC, Random Forest (RFC), and Naive Bayes, achieving a 0.89 F1-score on a Nepali news corpus. Applied TF-IDF summarization for automatic headline generation and sentiment grouping.
View project
Live

Vidhi-AI: Legal RAG & Ingestion Engine

Python, FastAPI, Qdrant, Celery, RabbitMQ, Redis, Selenium, Tesseract OCR, Docker

A retrieval-augmented generation (RAG) and legal ingestion engine for Nepalese legal documents. A Selenium scraper downloads PDFs from the Nepal Law Commission website, a parsing engine extracts text from mixed Nepali/English PDFs with a Tesseract OCR fallback, and a FastAPI vector search API serves results from Qdrant. Celery and RabbitMQ handle background ingestion jobs.
View project
Live

USMLE First-Aid RAG Chatbot

Python, LangChain, Qdrant, Hugging Face, Mixtral-8x7B, Docker

A retrieval-augmented generation (RAG) chatbot that answers questions from USMLE First Aid study material. Built with LangChain, Qdrant as the vector database, Hugging Face embeddings, and Mixtral-8x7B-Instruct as the LLM, with Qdrant running in Docker.
View project
Live

GPS Tracking System

PHP, JavaScript, MySQL, Composer, eSewa API

A PHP-based GPS tracking system with separate admin, manager and user panels, and eSewa payment integration with success and failure handling.
View project
Live

traceviz

Python, FastAPI, SQLite, React, TypeScript, D3, Docker

A visual traceroute and MTR-style route monitor. Each hop is plotted on a world map as it is discovered, with per-hop loss, latency and jitter, and a summary of where latency and packet loss start. It runs from the terminal (CLI) and a browser (web UI), with trace history saved to SQLite. Geolocation uses bundled DB-IP Lite databases, so no IP lookup API is called. Everything ships as one Docker image.
View project