AI research
Low-resource NLP · Speech · RAG evaluation · Multilingual embeddings · Dataset construction · PyTorch · TensorFlow · Hugging Face
About Samir
Research discipline. Engineering range. Founder urgency.
I’m Samir Wagle, a computer engineer and freelance AI engineer based in Lalitpur, Nepal, in the Kathmandu Valley. I build AI automation, retrieval systems, and production software for founders and teams, with remote collaboration available worldwide.
01
Research lab to production floor.
Roles overlap because my work does: research, client systems, product engineering, operations, and technical founding.
Full-stack and platform engineering
Fintech company of NepalMobile, web, Go and Python services, PostgreSQL, CI/CD, Linux operations, AI-assisted identity workflows, and architecture.
Freelance AI Engineer
Voice, RAG, and agentic systemsASR to LLM to TTS agents, retrieval systems, telephony and CRM integrations, automation workflows, and deployment infrastructure.
Speech AI Engineer
Mantra IdeasNepali and multilingual ASR/TTS adaptation, speech-data preparation, forced alignment, augmentation, and low-latency inference.
Research and Development Engineer
CityTechEmbedded C and Android software for payment terminals, TCP/MQTT paths, Android migration, NFC, card-reader, and PIN-pad integration.
Undergraduate Researcher
KU Information and Language Processing LabNepali sentiment and offensiveness models, dataset creation, multilingual embeddings, and English–Nepali–Tamang machine translation.
Co-Founder, Tech Lead
Impact Nexus NepalTechnical operations, identity and access administration, nonprofit technology programs, grants, and organizational infrastructure.
Public Relations and Social Media
Kathmandu University Literature AcademyPublic identity, digital presence, website and email operations, campaign graphics, posters, and social content.
02
Confidence arrives one boundary at a time.
How I decide, measure, migrate, and state limits.
- 01
Start with evidence.
Read source, logs, data, and real user paths before prescribing architecture.
- 02
Keep simple baselines alive.
A stronger model must beat a cheaper, inspectable alternative on the metric that matters.
- 03
Preserve contracts during change.
Migration succeeds when downstream systems do not have to move at the same time.
- 04
State what is not proven.
Built, tested, deployed, and effective are different claims. I label them separately.
03
Range without making tools the story.
Tools change. Ownership across research, application, infrastructure, and communication stays useful.
Automation and agents
n8n · LLM agents · Webhooks · Vector retrieval · Model guardrails · Human approval gates
Product engineering
Go · Python · TypeScript · React · Next.js · Node.js · FastAPI · Flutter · PostgreSQL
Systems and hardware
Linux · Docker · CI/CD · Servers · Embedded C · Android · TCP/IP · MQTT · POS devices
Design and communication
Figma · Photoshop · Illustrator · Technical writing · Photography · Video · Public communication
04
Computer engineering, grounded in language research.
BE, Computer Engineering
Kathmandu University · 2020–2025
GPA 3.72 / 4.0+2 / Intermediate
Capital Secondary School · 2018–2020
Koteshwor, KathmanduPrimary and secondary education
Adarsha Vidya Mandir Madhyamik Vidyalaya · 2008–2018
“Wagle has been an outstanding collaborator since day one, demonstrating a strong work ethic, adaptability, and a sharp problem-solving mindset. His ability to seamlessly integrate mathematics, programming, and research insights has been instrumental in our projects.”
05
Code, models, writing, and professional record.
Research focuses on technologies underserved languages are usually forced to borrow from English. Production work spans financial software, distributed services, products, infrastructure, compliance, and automation.
Talk