About
I am a Staff Software Engineer specializing in AI-native platforms, distributed systems, and LLM-based agentic workflows. I design and scale reliable systems that balance performance, cost, and developer experience.
Currently serving as a Staff Software Engineer at Fusemachines, I focus on building AI-native workflow platforms, vector search systems, and resilient distributed job orchestration for enterprise use cases.
Over 10+ years, I have led cross-functional teams, owned end-to-end system design, and delivered measurable improvements in reliability, latency, operational costs, and developer workflows across healthcare, EdTech, and enterprise platforms.
I enjoy working at the intersection of product and engineering, shipping practical LLM features with strong evaluation, observability, and production readiness.
Beyond the Code
When I am not building software, I recharge through a few hobbies I genuinely enjoy:
Trekking & Travelling
Multi-day treks, hiking trails, and exploring new places and cultures
Photography
Portraits, landscapes, and street photography
Acoustic Guitar
Playing rhythm parts and practicing fingering on acoustic guitar
Basketball
Following the NBA and Kawhi Leonard - playing pickup games when I can
Gaming
Casual PC and mobile gaming to unwind
By the Numbers
Core Technical Arsenal
AI & Languages
Data & Backend
Cloud, DevOps & Observability
Resume
Professional Summary
Umesh Subba Limbu
Staff Software Engineer with 10+ years of experience designing, building, and scaling distributed systems and AI-native platforms. Specialized in LLM-based agentic workflows, advanced RAG architectures, and production engineering. Proven track record of leading cross-functional teams of 10–15 engineers, owning system design end-to-end, and delivering practical improvements in system reliability, latency, operational costs, and developer workflows.
Contact Information
Email: umes4ever@gmail.com
Education & Credentials
Bachelor of Engineering (B.E.) in Computer Science
May 2015
NMAM Institute of Technology (NMAMIT), India
(CGPA: 9.42/10.0)
M320: Data Modeling
Feb 2021
MongoDB University
Professional Experience
Staff Software Engineer
Nov 2025 – Present
Fusemachines Nepal Pvt. Ltd., Kathmandu
- Designed and built a vector memory layer using pgvector, improving semantic retrieval and reducing redundant LLM context token usage by 25–30%.
- Built a distributed workflow orchestration layer using Python Procrastinate to handle thousands of async jobs daily with built-in retries, failure recovery, and logging.
- Integrated Playwright end-to-end testing into CI/CD pipelines, reducing production regressions by 35% and improving release predictability.
- Reduced critical path latency across AI workflows by 20% through contextual caching, retrieval optimization, and prompt refactoring.
Assistant Technology Manager / Project Lead
Jan 2024 – Oct 2025
Fusemachines Nepal Pvt. Ltd., Kathmandu
- Led cross-functional teams of 10–15 engineers, taking multiple enterprise platforms from initial design through to production.
- Designed scalable, HIPAA/GDPR-compliant backend systems to ensure secure and auditable handling of regulated enterprise data.
- Reduced deployment-related production incidents by 25–30% by improving CI/CD health checks, code review guidelines, and release governance.
- Managed system design reviews for distributed microservices, focusing on service boundaries, asynchronous communication, and data schemas.
- Lowered AWS infrastructure costs by ~15% via infrastructure right-sizing, compute instance scheduling, and query optimization.
Project Experience
Agentic OS for Businesses
2025 – Present
AI-native workflow automation platform using LLM agents for enterprise operations
- Built an AI-native workflow automation platform using LLM agents to orchestrate complex enterprise operations.
- Designed a pgvector-based memory system enabling semantic retrieval and reducing redundant context usage by ~20–30%.
- Implemented a distributed workflow execution system handling thousands of jobs per day with reliability guarantees.
- Improved system stability via Playwright-based testing, reducing production regressions by ~25–40%.
- Reduced LLM token consumption by ~15–25% through conversation summarization and context optimization strategies.
AI Studio
2024 – 2025
Enterprise AI development platform enhancements
- Improved deployment pipelines, reducing manual deployment overhead and improving release cadence.
- Built a real-time automation system using Python watchdog for file-based workflow triggers.
- Implemented an observability stack, reducing incident diagnosis time by ~20–30%.
- Enhanced Kubernetes deployment workflows, improving rollout stability and service reliability.
Healthcare Platform
2022 – Present
Enterprise healthcare system with compliance requirements
- Designed HIPAA/GDPR-compliant system architecture for regulated enterprise workflows.
- Supported systems handling sensitive enterprise data with strict security and audit requirements.
- Improved engineering governance for reliability, compliance, and production readiness.
Online Learning Platform
2019 – 2021
Large-scale education platform
- Led 10+ engineers across the full product lifecycle and Agile delivery.
- Delivered 30+ stable production releases with consistent deployment execution.
- Resolved 200+ production issues, improving system reliability and user experience.
- Supported systems serving thousands of active users in live learning environments.
- Improved production stability through structured engineering and release discipline.