AI Engineer AI Researcher
Building reliable NLP, LLM, and responsible AI systems.
Published on arXiv AWS Certified Solutions Architect Author of MemBlocks on PyPI
I'm Ashish Pandey, an AI engineer and AI researcher based in Patan, Lalitpur, Nepal. I work on building and studying language based AI systems, with a focus on making them more reliable, fair, and usable in real-world settings.
At CAIR-Nepal, I work on understanding and evaluating sociocultural bias in large language models for the Nepali language, where I helped design a structured evaluation framework and curated datasets for analysis across multiple model families.
I'm currently pursuing Computer Engineering at Pulchowk Campus, IOE Nepal, and hold an AWS Solutions Architect – Associate certification.
My work spans the full ML lifecycle from model development with PyTorch and TensorFlow to building production systems using FastAPI, Django, and LangChain, deployed on AWS with Docker. I've also built LLM memory systems, multi-agent applications, and voice-driven AI tools.
AI Research Intern
CAIR-Nepal (Center for AI Research Nepal) Present
Researching and evaluating sociocultural bias in large language models for the Nepali language. Helped design a structured evaluation framework and curated datasets for analysis across multiple model families.
Bachelor of Computer Engineering
Pulchowk Campus, Institute of Engineering (IOE), Tribhuvan University In progress
Computer engineering coursework alongside applied work in machine learning, natural language processing, and production AI systems.
AWS Certified Solutions Architect – Associate
Amazon Web Services Certified
Cloud architecture certification applied to deploying ML services on AWS with Docker, FastAPI, and Django.
The languages, frameworks, and platforms Ashish Pandey uses for machine learning research and production AI engineering.
A collection of things I've built, experiments I've run, and problems I've enjoyed solving. Browse all AI projects by Ashish Pandey.
Featured · Open source
A Python library for modular, section-aware memory in LLM applications. Instead of flat RAG or dumping whole histories, MemBlocks organises memory into independent, attachable blocks, each with its own storage, retrieval, and update strategy, backed by Qdrant and shipped with a FastAPI demo and an MCP server.
AI-powered meeting management app with multi-agent scheduling and email automation.
Voice-controlled system automation agent enabling hands-free control, email, and browser automation.
Assistive app for the deaf and hard of hearing with ML-powered environmental sound detection.
Real-time ASL recognition integrated into video calls with live transcription to English.
Peer-reviewed and preprint research by Ashish Pandey in AI, NLP, responsible machine learning, and large language model evaluation. See all publications.
Under Review · PLOS Digital Health
Neupane, H. K., Mishra, B. K., Pandey, A., Poudel, S., & Simkhada, P.
A framework and roadmap for responsible artificial intelligence in the health systems of low- and middle-income countries, developed from the Nepali context. Co-authored by Ashish Pandey; under review at PLOS Digital Health.
Preprint
Pandey, A., & Chhetri, T. R.
A dual-metric framework for evaluating social bias in large language models, with evidence from an underrepresented Nepali cultural context. Ashish Pandey and T. R. Chhetri introduce a structured evaluation framework and curated datasets to analyze bias across multiple LLM families.
Quick answers about Ashish Pandey, his AI research, and his engineering work.
Open to research collaborations, AI roles, and interesting problems.
ashishpandey2062@gmail.com
Location
Patan, Lalitpur, Nepal