Buddhi Sagar Poudel

// Who I am

More than
the résumé.

I grew up curious about how things work — not just on the surface, but at the level of systems, patterns, and the invisible logic underneath. That curiosity eventually found a home in engineering.

Today I'm a Data Engineer and AI Engineer based in Kathmandu, Nepal, working on problems that matter: healthcare data routing, real-time fraud detection, intelligent farming, and the kind of infrastructure that makes all of it possible.

I hold a Bachelor's in Computer Science and Engineering from Nepal Engineering College, and I've spent the years since building things — pipelines that move billions of rows, models that catch fraud before it happens, systems that speak back in natural language.

I believe in understanding deeply before building fast. That's shaped how I approach every system I touch — from contributing a Prometheus fix to a 60k-star open-source project, to designing a multi-threaded ETL engine with custom dependency graphs at ExtensoData.

When I'm not engineering, I'm usually reading research papers, following the frontier of LLMs and data systems, or thinking about how technology can be genuinely useful to people in Nepal and beyond.

Nepal 🇳🇵 B.E. CSE · Nepal Engineering College IEEE Published Fusemachines AI Fellow HackerRank SQL Advanced Open to Remote / Relocation

// Career path

The
Journey.

Sep 2024 – Nov 2024

Data Engineer

ExtensoData Pvt. Ltd.

Apache Iceberg migration, multi-threaded ETL engine, Kafka real-time pipelines, AML anomaly detection, RAG systems

May – Sep 2024

Data Engineering Intern

F1Soft International · Nepal's leading fintech

ETL on Microsoft Fabric & Azure, Power BI dashboards, OLAP dimensional modeling

May – Oct 2024

AI Fellowship

Fusemachines Nepal

Microdegree in AI · LLMs, Computer Vision, NLP, MLOps, Generative AI

Jun 2023 – Dec 2024

Freelance Engineer

Independent

AWS ML pipeline (1B+ rows), GCP streaming analytics, full-stack builds

Oct 2019 – Oct 2024

B.E. Computer Science & Engineering

Nepal Engineering College, Bhaktapur

DSA, AI, Big Data, Cloud Computing, NLP, Probability & Statistics

// Technical profile

What I Build

The domains where I spend most of my engineering energy — each one backed by real production work.

⚡

Data Engineering

I build the pipes. Multi-threaded ETL engines with config-driven dependency graphs, Apache Iceberg migrations, real-time Kafka pipelines, and HDFS data aging modules. At ExtensoData I designed the orchestration layer from scratch. At F1Soft I automated ingestion on Microsoft Fabric and Azure.

🧠

AI & Machine Learning

From anomaly detection engines in production AML systems to LLM fine-tuning, RAG pipelines, and Text-to-SQL systems. I've built and deployed models for fraud detection, merchant segmentation via NLP, plant pathogen classification, and underwater trash detection (published in IEEE).

☁️

Cloud Platforms

AWS (Glue, S3, SageMaker, Athena, Redshift, Bedrock), GCP (BigQuery, Pub/Sub, Dataflow, Vertex AI, Looker), Azure (Fabric, Data Factory). Built a pipeline processing 1B+ rows on AWS for a real client. Streaming analytics on GCP for e-commerce user behavior.

🔬

LLMs & Generative AI

Building with LangChain, RAG architectures, MCP-integrated tool agents, and Text-to-SQL systems. At ExtensoData I built a multi-document RAG pipeline with dynamic document selection. At Maitri, working with Agentic AI for healthcare EDI workflows.

🔐

Systems & Security

Cryptographic e-voting with blind signatures and zero-knowledge proofs. Steganographic data masking. Persistent-token secure email. Open source contributor to MinIO — fixed a Prometheus histogram metric incompatibility in the 60k-star production codebase.

🏗️

Full-Stack Engineering

End-to-end platforms with proper architecture: RBAC e-commerce with three portals, SMTP-based secure mail services, cryptographic voting systems, ML-integrated mobile banking fraud detection, and a smart farming platform deployed for Nepali farmers.

// Stack

Technical Skills

Language Breakdown

Based on GitHub repo analysis

// Work history

Professional Experience

Software Engineer

Maitri Holdings Services Pvt. Ltd.

Nov 2025 – Present

Engineering EDI file routing architectures for US healthcare claims processing
Analyzing health insurance eligibility, claims, and enrollment data for accuracy and compliance
Building Agentic AI workflows with Scala and Apache NiFi for automated data pipeline orchestration
Maximizing insurance data efficiency to improve customer outcomes in the US healthcare domain
Contributed to a large-scale chart processing engine for PDF and XML healthcare documents
Reduced compute costs by ~60% through pipeline optimization and scaling improvements
Worked with OCR engines and pipelines including DocTR and LightOnOCR
Built LOINC and ICD code extraction workflows using custom LLM-based engines
Built AI agents and automated agentic workflows for production support and operational tasks

Data Engineer

ExtensoData Pvt. Ltd.

Sep – Nov 2024

Led research and migration of existing data infrastructure to Apache Iceberg, significantly improving performance, scalability, and reliability for large-scale processing
Designed and built a config-driven multi-threaded ETL engine with parallel execution and dependency management — significantly faster processing with flexible orchestration
Built a custom anomaly detection engine for identifying unusual patterns in financial transactions; deployed it in production. Contributed to AML chain detection and feature weight modeling
Built a RAG pipeline retrieving from multiple documents with dynamic document selection; and a Text-to-SQL system returning natural language responses
Developed Kafka real-time pipelines for loyalty points processing, SCD Type-2 MDM, and data lineage tracking
Mentored new hires on data ecosystems, best practices, and technologies

Data Engineering Intern

F1Soft International Pvt. Ltd. — Nepal's leading fintech

May – Sep 2024

Built ETL pipelines on Microsoft Fabric & Azure with automated ingestion and transformation workflows
Implemented ML-driven data analysis and surfaced insights via Power BI dashboards
Assisted in dimensional modeling and OLAP data warehouse design

Freelance Engineer

Independent

Jun 2023 – Dec 2024

Ride Fare Prediction · AWS: End-to-end ETL + ML pipeline processing 1+ billion rows with AWS Glue and S3. Deployed predictive models in SageMaker for optimal fare forecasting
Streaming User Behavior Analytics · GCP: Near-real-time pipeline using Pub/Sub, Dataflow, and BigQuery for an e-commerce startup. Raw → gold data layers with interactive Looker Studio dashboards

// What I've built

Featured Projects

Systems I've designed, engineered, and shipped — each one solving a real problem.

🔬

✓ Merged · Open Source

MinIO — Prometheus Observability Fix

60,400 ⭐ · Production-grade distributed object storage · Used by Netflix, Dropbox, Cloudflare

Identified that ttfb_seconds_distribution was incorrectly typed as a gauge, making it incompatible with Prometheus' histogram_quantile() — the function every SRE team relies on for p50/p95/p99 latency SLOs. Changed metric type to histogram, updated the descriptor accordingly. Reviewed and approved by harshavardhana (MinIO's lead maintainer) and merged into master in April 2025, then cherry-picked into the stable release.

✓ Merged · PR #20999 · minio/minio ● Open · PR #4225 · delta-io/delta

01 — Featured

OTG — Mobile Banking
Fraud Detection Engine

⛓ otg-fd + OTGFD-ML-Server

A production-grade fraud detection system simulating complete mobile banking flows — transfers, balance queries, payment histories — with an ML anomaly engine layered on top. Uses isolation forests and autoencoder reconstruction error thresholds to detect suspicious behavioral patterns in real time without requiring labeled fraud data. Confidence scores and anomaly explanations surface directly in the UI.

PythonFlaskAutoencodersIsolation ForestReactJavaScriptREST API

Frontend ↗ ML Server ↗

🏦

OTG Banking Platform

Real-time behavioral anomaly detection · Mobile banking simulation

Anomaly Detection Real-time Alerts

02 — Featured

SAGE — Smart Agriculture
Intelligence Platform

⛓ smart-farming-solution + fusemachines_SmartFarmingSolutions

Project SAGE — a comprehensive AI platform modernizing Nepali agriculture. Multilingual (English + Nepali) crop recommendation engine analyzing soil NPK, temperature, and rainfall. AI-powered pathogen detector: upload a leaf photo, get disease classification and treatment recommendations instantly. Real-time soil moisture prediction for precision irrigation.

PythonTensorFlowYOLOCNNscikit-learnAWSMultilingual UI

▶ Watch Demo GitHub ↗

03 — Featured

SecureInbox —
Persistent Temp Mail

⛓ secure-mail-server + secureinbox-client

Temporary email with a key differentiator: session-persistent browser tokens. Generate a secure token, save it to your browser. Return days later — your inbox is still there, full message history intact. The backend implements full SMTP reception, per-address token issuance, TTL-based expiry, and encrypted session storage.

Node.jsSMTPReactToken AuthIndexedDBEncryption

Server ↗ Client ↗

📬

SecureInbox

Your temp mailbox survives the session — via persistent browser tokens

SMTP Token Persistence

04 — Featured

Nexus Commerce —
Multi-Portal E-Commerce

Full e-commerce ecosystem with three independent role-based portals: customer storefront, seller dashboard, and admin control panel. Fine-grained RBAC across all layers — product management, inventory, order processing, and platform analytics. JWT with refresh token rotation, integrated payment gateway, seller onboarding workflows.

Node.jsReactPostgreSQLRBACJWTPayment GatewayREST API

GitHub ↗

🛍️

Multi-Portal Commerce

Customer · Seller · Admin · RBAC · Payment integration

3 Portals RBAC

05 — Featured

CipherVote —
Cryptographic E-Voting

⛓ evoting-api + evoting-client

A mathematically rigorous, breach-resistant voting platform. Votes are encrypted end-to-end — the server never sees plaintext ballots. Implements blind signatures, zero-knowledge proof of vote validity, and public audit-trail generation. Designed to resist replay attacks, double voting, insider threats, and ballot stuffing.

Node.jsCryptographyBlind SignaturesZKPReactJavaScript

API ↗ Client ↗

🔐

CipherVote

Blind signatures · Zero-knowledge proofs · Tamper-proof ballots

ZKP Blind Sig

// More work

Everything Else

🎭Python + JS

DataMask

Steganography tool — hides sensitive data (PDFs, text, credentials) inside ordinary image files using LSB encoding. Invisible to inspection.

PythonSteganographyReact

📊JavaScript

Axis Platform

Interactive analytics and data visualization platform for growth-oriented digital strategy. Blog-integrated, dynamic dashboard. Live on Vercel.

ReactD3.jsVercel

🚕Python / Jupyter

NYC Taxi · AWS Pipeline

End-to-end DE + DS + MLOps pipeline on AWS. Ingest → transform → train → deploy. Covers the complete data engineering to ML stack on cloud.

AWS GlueS3SageMaker

🏍️JavaScript

BikeRent

P2P bike rental with Khalti payment integration and Google Maps for real-time location. Private owners list bikes; renters find nearby options.

Node.jsKhaltiGoogle Maps

🔄Python

SyncFiles

Automatic LAN filesystem sync over SFTP. Keeps two directories in sync across devices on the same network — no cloud, no manual setup.

PythonSFTPNetworking

🌬️Python

Airflow Pipelines

Production-style Apache Airflow DAG orchestration with scheduled batch jobs, retry logic, SLA monitoring, and alerting hooks.

AirflowPythonPostgreSQL

// Research

IEEE Publication

Published · IEEE · Dec 2024

Towards Underwater Sustainability with AIoT: Underwater Trash Management System Comprised of YOLOv8 with IoT-Applied Segmentation

IEEE · December 25, 2024

Proposes an AIoT system for automated underwater trash detection and segmentation. Uses YOLOv8 for real-time object detection combined with IoT-enabled actuation for cleanup operations. Addresses ocean sustainability through applied computer vision and edge computing.

YOLOv8IoTComputer VisionSegmentationAIoT

More thanthe résumé.

TheJourney.