Utsav Paudel (also spelled Utsab Paudel, Utshab Paudel, Utsav Poudel) is a Machine Learning Engineer based in Kathmandu, Nepal, with 2.5+ years of experience building Generative AI systems, RAG pipelines, LLM applications, and Voice AI agents on AWS.

Who is a Machine Learning Engineer in Nepal?

Utsav Paudel is a Machine Learning Engineer in Nepal, currently working at Leapfrog Technology in Kathmandu. He specializes in Generative AI, RAG architectures, LLMs, and AWS-based AI solutions.

What does Utsav Paudel do?

Utsav Paudel builds production AI systems including RAG chatbots, Text-to-SQL engines, and Voice AI agents. He works with AWS Bedrock, Claude, LangChain, PyTorch, and FastAPI.

Utsav Paudel — Machine Learning Engineer | Generative AI, RAG & LLM Systems

§ 01 · About

A brief self-portrait, in three columns.

I bridge the theory with the production-grade.

Machine Learning Engineer with 2.5+ years building Generative AI, RAG architectures, and cloud-native AI for fintech and healthcare. My favourite systems are the ones that think in one modality and speak in another — a phone call that becomes a database row, a stack of PDFs that becomes a conversation.

Right now I'm at Leapfrog Technology, deep in AWS Bedrock, Claude, and Amazon Nova. Before that, Fusemachines and E.K. Solutions, where I cut my teeth on retrieval, prompt engineering, and the unromantic art of making transformers fit inside a budget.

Off-hours I live outside — trekking Langtang, following waterfalls, photographing the mountains that raised me. The thing I believe about engineering: you ship better systems if you remember how small you are.

I like working on problems that don't already have a clean answer. Ambiguity is where the interesting engineering lives. If that sounds like your weekend, let's talk.

Years in the trade

2.5+

Since late 2023

Documents retrieved

10,000+

Across RAG & GraphRAG pipelines

LLM extraction accuracy

95%

On structured extraction benchmarks

§ 02 · Selected work

Four systems, shipped.

Case 01 — Fintech / VisaAWS · Bedrock

Visa Analytics Chatbot.

A four-stage RAG pipeline on AWS Bedrock + Claude Haiku 4.5, with an ID-mapping citation strategy so every answer points to the exact paragraph in the source document. No more hallucinated references.

Role: Lead ML Engineer
Year: 2025 · Leapfrog
Stack: Bedrock, Claude, Python
Impact: Deployed to prod

AWS BedrockClaude Haiku 4.5RAGPythonFastAPI

pipeline.schema 4-stage

01retrievehybrid · bm25+dense

02re-rankcross-encoder · top-k

03cite-mapid → span

04answerclaude haiku 4.5

Case 02 — Fintech / LendingNova Pro · pgvector

GoLend AI Assistant.

A Text-to-SQL assistant powered by Amazon Nova Pro with BM25 + pgvector hybrid search, IVFFlat / HNSW indexing, MMR re-ranking, and AWS Guardrails — so loan officers can ask "which applicants missed their third installment?" and get a correct query back, every time.

Role: ML Engineer
Year: 2025 · Leapfrog
Stack: Nova Pro, Postgres, Python
Impact: 50% faster queries

Nova PropgvectorBM25HNSWPostgreSQL

nl → sql · preview accepted

-- "which applicants missed their third installment?"
SELECT a.id, a.name, a.email
FROM   applicants a
JOIN   installments i ON i.applicant_id = a.id
WHERE  i.sequence = 3
  AND  i.paid_at IS NULL
  AND  i.due_at < now()
ORDER BY i.due_at DESC
LIMIT 50;

Case 03 — Healthcare / VoiceNova Sonic · Twilio

Voice AI for patient engagement.

A real-time IVR agent that picks up the phone, confirms insurance, and schedules follow-ups — built on Amazon Nova Sonic + Twilio, with a Textract → Nova Pro → OpenEMR pipeline behind it. 60% reduction in scheduling overhead.

Role: ML Engineer
Year: 2025 · Leapfrog
Stack: Nova Sonic, Twilio, Lambda
Impact: 60% less overhead

Nova SonicTwilioTextractOpenEMRAWS Lambda

call · transcript · 00:23 live

caller Hi, I need to move my follow-up, my daughter's sick.

agent Of course. Dr. Mehta has openings Thursday at 10 or Friday at 2 — which works?

caller Friday at 2, please.

agent Booked. I'll text the confirmation to the number on file.

Case 04 — Fintech / AnalyticsNeo4j · GraphRAG

Lender IQ · GraphRAG.

A knowledge-graph-backed retrieval system over 10,000+ financial documents. Entities and relationships extracted with LLMs, stored in Neo4j, and queried with GraphRAG to answer questions traditional RAG can't — "which guarantors are co-signing loans across three or more portfolios?"

Role: ML Engineer
Year: 2024 · Fusemachines
Stack: Neo4j, LangChain, Python
Impact: 40% accuracy lift

Neo4jGraphRAGLangChainFastAPIGPT

knowledge-graph · subset neo4j

§ 03 · Career

The ledger, in reverse chronology.

May 2025
Present

Machine Learning EngineerLeapfrog Technology Pvt. Ltd.

Visa Analytics Chatbot on AWS Bedrock + Claude Haiku 4.5 — 4-stage RAG, ID-mapping citations.
GoLend AI: Text-to-SQL on Amazon Nova Pro + PostgreSQL pgvector.
Voice AI agent on Nova Sonic + Twilio — 60% less scheduling overhead.
Automated patient-referral pipeline, Textract → Nova Pro → OpenEMR.

BedrockNova ProNova SonicTwilioLambda

Sep 2024
May 2025

Associate ML EngineerFusemachines Pvt. Ltd.

RAG & GraphRAG systems with 40% accuracy improvement across 10,000+ financial documents.
LLM extraction pipelines (GPT, Gemini) reaching 95% accuracy.
Retrieval optimization delivering 50% query time reduction.

Neo4jLangChainGPTGeminiPython

Jan 2024
Jun 2024

Associate ML EngineerE.K. Solutions Pvt. Ltd.

5+ scalable ML APIs on FastAPI + PyTorch serving 1,000+ daily requests.
25% LLM accuracy lift via Chain-of-Thought prompt engineering.
35% transformer performance optimization with 20% memory reduction.

FastAPIPyTorchDockerRedis

Oct 2023
Jan 2024

ML Engineer TraineeE.K. Solutions Pvt. Ltd.

Foundational ML model development and training pipelines.
Prototyping of NLP and computer-vision solutions.

PyTorchOpenCVNumPy

§ 04 · Toolkit

The workshop, laid out flat.

AI · ML frameworks07

Where the thinking happens.

PyTorch
LangChain
LangGraph
LangSmith
CrewAI
Google ADK
OpenCV

Languages03

Daily drivers.

Python
SQL
C++

Web frameworks03

APIs & edges.

FastAPI
Django
Django REST

Databases04

Where state lives.

PostgreSQL (pgvector)
Neo4j
MongoDB
SQLite

Cloud06

AWS, mostly.

AWS Bedrock
Lambda
EC2
SQS
SNS
Vertex AI

Infrastructure05

Plumbing, done right.

Docker
MLFlow
Redis
Celery
Alembic

Utsav builds machines that think, speak, & reason.

Visa Analytics Chatbot.

GoLend AI Assistant.

Voice AI for patient engagement.

Lender IQ · GraphRAG.

Utsav builds

machines that

think, speak,

& reason.