Contact
Phone
+977 9860334317
aadarshapandit@gmail.com
Address
Kathmandu, Nepal
Website
aadarshapandit.com.npEducation
BTech in Artificial Intelligence
Kathmandu University
2024 – Present
GPA: 3.95 / 4.0
Science (+2)
Kathmandu Bernhardt College
2022 - 2024
Skills
Programming
Python, JavaScript, TypeScript, SQL
AI / ML
PyTorch, TensorFlow, Scikit-Learn, HuggingFace, Random Forest, Pandas, NumPy, NLP, ASR
Web & Backend
React, Next.js, Node.js, FastAPI, WordPress
Mobile Development
Flutter
Tools & Platform
Git, Docker, Linux, Blender
Language
NepaliNative
EnglishProfessional
HindiConversational
AADARSH
PANDIT
AI & Web Developer
BTech AI student at Kathmandu University. Specializes in low-resource NLP for Nepali (ASR, code-switching) and ships production-grade ML systems alongside high-performance creative web experiences.
Professional Experience
Feb 2026 – Present
AI Lead
Contract- Delivered 3+ client-commissioned AI projects including custom chatbots (with WordPress plugin integration), domain-specific predictors, and fine-tuned LLMs.
- Lead model fine-tuning pipelines end-to-end: data curation, training, evaluation, and production handoff for business clients.
Jun 2025 – Nov 2025
AI Researcher — Nepali ASR
Research PaperIndependent Research
- Developed a reward-guided fine-tuning framework for Whisper, utilizing a Random Forest model (81% accuracy) to filter noisy samples from a 70+ hour Nepali speech dataset.
- Achieved an 11–12% relative improvement in transcription accuracy, reducing WER from 5.55% to 4.89% and CER from 5.04% to 4.52%.
Jan 2026
Web Developer
Contract- Built and deployed corporate website using WordPress and Elementor with custom page templates and responsive layouts.
- Configured hosting, SEO metadata, and performance settings achieving fast page loads for a business-focused audience.
Projects
Feb 2026
Aadarsh Travels
Next.js, UI/UX- Architected luxury, high-performance web experience showcasing travel routes using Next.js.
- Engineered custom scroll-synced 4K WebP frame animation system (253 frames), conditionally rendering mobile sizes to cut weight by 83%.
Jan 2026
Nepal Cricket Predictor
Scikit-Learn, FastAPI- Deployed a real-time predictive dashboard (FastAPI + React) using opponent, venue, and toss as features; model achieves 85% prediction accuracy.
- Trained a Random Forest classifier on historical T20I and ODI data, surface-level and match-condition features for Nepal national team fixtures.
Publications & Research
Under Review
Reward-Guided Fine-Tuning of Whisper for Low-Resource Nepali ASR
Random Forest reward model · 70+ hr dataset · 11–12% relative WER improvement
GitHub:Aadarshttech
LinkedIn:Aadarsh-Pandit