Utsabi Dangol
Data Engineer · ML Researcher · Full-Stack Developer
Python · Snowflake · Airflow · PyTorch · LLMs · React
I turn messy data into reliable pipelines and research ideas into working AI systems. With 4+ years of professional experience, I bring both engineering discipline and research depth from Snowflake-scale ELT to LLM-powered RAG systems and uncertainty-aware Reinforcement Learning.
What I work with
Technical Skills
A toolkit built across 4+ years of production software engineering and graduate-level ML research.
Data & ETL
Data Platforms
Cloud
Analytics & BI
AI & ML
Web Dev
Selected work
Projects
Policy Shaping with Uncertainty-Aware LLM for Multi-Task RL
Aug – Dec 2025
BERT-based LLM guidance combined with PPO using MC Dropout for uncertainty estimation. Achieved 99.2% success rate and 2009.96 reward AUC, outperforming Q-learning and DQN baselines.
LLM-Based PDF Question Answering using RAG
Feb 2026
RAG pipeline for interactive PDF QA using recursive text chunking, Watsonx embeddings, and ChromaDB. Gradio interface for document upload and similarity-based retrieval.
E-Commerce Analytics: Product Success & Price Prediction
Aug – Dec 2025
82.98% classification accuracy for product ratings using Amazon Reviews metadata with SHAP interpretability. RAG-based price prediction with FAISS & BM25 hybrid retrieval.
Detecting Bias in Social Networks & News
Jan 2026
Transformer pipeline for media bias detection — fine-tuned BERT, RoBERTa, ELECTRA, GPT-2 on MBIB dataset. ELECTRA achieved best F1 of 0.769.
Fault-Tolerant PostgreSQL with Raft Consensus
Nov 2024
Priority-based Raft consensus algorithm reducing leader election latency in distributed PostgreSQL clusters across 3–7 node Azure deployments.
Where I have worked
Experience
Data Engineer
LIS Nepal Pvt. Ltd.
- Led offshore team of 9 delivering Python/Airflow/Snowflake ELT pipelines, improving runtimes by 40% and costs by 30%
- Integrated 10+ data sources into Snowflake and GCP, enabling cross-channel attribution and ML-ready analytics
- Architected data marts and automated quality checks, reducing pipeline failures by 35% supporting 50+ dashboards
- Implemented RBAC and PII masking to secure 100% of sensitive customer fields, ensuring data privacy compliance
- Mentored 25+ junior engineers on best practices, improving code quality and delivery velocity
Junior Software Engineer
Impacters
- Built edge AI system (Project Saathi) using CNNs and transfer learning for real-time crop health monitoring
- Deployed ML models on solar-powered edge devices with offline capability for resource-constrained hardware
Associate Software Engineer
Asterdio Inc.
- Implemented customer segmentation using clustering algorithms to identify patterns and inform product strategy
- Optimized Node.js/MongoDB data pipelines and built analytics dashboards, improving query performance by 40%
Leadership
Entrepreneurial Lead
NSF I-Corps Hub Great Plains
- Conducted 40 customer discovery interviews to refine AI product concepts including legal research assistant (LLMs) and music transcription tool
- Applied lean startup methodology to validate problem-solution fit across both cohorts
Professional Development Chair
Rotaract Club of NewRoad City
- Organized and led professional development programs and workshops for club members
- Fostered skill-building initiatives to support members' career growth and community engagement