Natural Language Processing
Core NLP research spanning text generation, classification, and representation learning — with emphasis on clinical and multilingual settings across Japanese, Nepali, and English.
Santosh Gaire Sharma
AI Engineer & NLP Researcher
Osaka, Japan · Open to collaborations in Japan
Bridging Academic Research & Production-Ready AI | Agentic AI, LLMs, Healthcare NLP
Research Focus
About My Research
I am Santosh Gaire Sharma, an AI Engineer and NLP Researcher at Doshisha University, currently living in Osaka, Japan, and working in the Co-Creation Informatics Laboratory (CCILAB) under Prof. Shirahama. My research centers on NLP, low-resource languages, multilingual models, medical AI, and AI safety — from Japanese–Nepali clinical summarization and diffusion-based sensing to LLM-as-a-Judge evaluation for reliable agents. With industry experience building production multi-agent systems and products like GharKatha (gharkatha.com), I combine research rigor with real-world scalability. I am actively seeking PhD opportunities and research collaborations in Japan across AI/ML, NLP, and data science at the intersection of healthcare and intelligent systems.
Outside research, I love playing football, watching movies and series, and following the beautiful game — a big-time Manchester City supporter.
Low-Resource Languages
Building NLP systems for under-resourced language pairs such as Japanese–Nepali, including corpus cleaning, synthetic data, and evaluation for clinical and general domains.
Multilingual Models
Fine-tuning and adapting multilingual transformers (e.g. mT5) for cross-lingual generation, summarization, and transfer across Japanese, Nepali, and English.
Medical AI
Clinical NLP and healthcare sensing — medical text generation, cross-lingual clinical summarization, and diffusion models for robust multimodal health-related activity recognition.
AI Safety
Reliability and safety of AI systems: LLM-as-a-Judge evaluation pipelines, agent benchmarking, and detection of AI-generated text for integrity and information security.
Agentic AI & LLMs
Multi-agent orchestration, MCP-based tool use, RAG pipelines, and reasoning systems that move research ideas into production-grade applications.
Academic Background
Education
Sep 2025 – Sep 2027
Master's Program
Information and Computer Science
Doshisha University · Kyoto, Japan
Co-Creation Informatics Laboratory (CCILAB), Prof. Shirahama
Thesis / Research
Human activity recognition via diffusion-based missing modality imputation; Cross-lingual Japanese–Nepali medical text summarization
Key Coursework
Statistical Machine LearningPattern RecognitionResearch and EthicsNon-linear PhysicsNeuroscienceNature-Inspired ComputingPresentation and Documentation PreparationResearch and Development
Apr 2019 – Apr 2024
Bachelor of Engineering
Computer Engineering
Tribhuvan University · Nepal
Thesis / Research
Refinetograph: A Machine Learning Approach Toward Image Enhancement
Key Coursework
Artificial IntelligenceDatabase SystemsSoftware EngineeringComputer Networks
Career Path
Experience
Research Assistant
Doshisha University · Kyoto, Japan
ResearchMay 2026 – Present
- Researching human activity recognition by imputing missing modalities with a diffusion model
- Developing cross-lingual summarization systems for Japanese–Nepali medical text generation
Student Tutor
Doshisha University · Kyotanabe, Japan
ResearchApr 2026 – Present
- Mentoring students in specialized AI and computer science fields
- Guiding development of proprietary research systems and assisting in technical paper composition
AI Engineer
SoranoAI · San Francisco, CA
IndustrySep 2025 – Apr 2026
- Architected framework-agnostic "LLM-as-a-Judge" evaluation pipelines to benchmark reliability and safety of AI agents
- Designed and deployed multi-agent workflows and LLM-based reasoning systems for predictive product features
- Bridged AI research with commercial productization — robust frontend, backend routing, and Stripe payment integration
- Drove 20%+ paid conversion rate within three weeks of beta launch
Software Engineer
TeamOne Technologies · Mahalaxmi, Nepal
IndustryMay 2024 – Dec 2024
- Engineered scalable, high-performance web applications with Next.js and Node.js
- Deployed interactive, low-latency live mapping systems using Leaflet.js and React-Leaflet
- Collaborated across cross-functional teams to optimize code performance and deliver production-grade software
Technical Proficiency
Skills
Research and data science first — engineering and infrastructure as supporting depth. Select a skill to highlight related work.
Research & NLP
Data Science & ML
Agentic AI & LLMs
Software Engineering
supportingInfrastructure & Tools
supportingSelect a skill to see related projects below.
Selected Work
Projects
Flagship research first. Click a project to highlight the skills behind it — or filter via Skills above.
ResearchTargeting COLING 2027
NepMedJP
Cross-lingual summarization system for generating Nepali medical text from Japanese clinical documents, targeting low-resource language pairs in healthcare. Active research with cleaned multilingual corpora and mT5 fine-tuning.
NLPData / ML
PythonmT5Hugging FaceTransformersNLP
ResearchPublication planned
Cross-Diffusion Model for HAR
Research on reconstructing missing sensor modalities in human activity recognition using cross-sensor diffusion techniques for robust multimodal sensing in health-related contexts.
Data / ML
PythonPyTorchDiffusion ModelsHAR
PublishedPublished · IJICTDC 2024
Refinetograph
Integrated web application combining SRGAN for super-resolution, Convolutional Autoencoder for denoising, and Zero-Reference Deep Curve Estimation for low-light image enhancement.
Data / MLEngineering
PythonPyTorchSRGANZeroDCEFlask
ResearchResearch prototype
Human vs AI Text Classification
CNN-based non-linear classifier distinguishing human-written text from AI-generated content for academic integrity and information security applications.
NLPData / ML
PythonCNNJupyterNLP
ProductionProduction deployment
LLM-as-a-Judge Evaluation Pipeline
Framework-agnostic evaluation system to benchmark reliability and safety of production AI agents using automated LLM judging, built during industry work at SoranoAI.
AgentsEngineering
PythonLLMsEvaluationAPI Design
ResearchAgent tooling study
MCP Agent Workflows
Hands-on exploration of the Model Context Protocol for connecting AI agents to external tools, data, and services in structured, extensible workflows.
AgentsEngineering
MCPPythonAgentsJupyter
ResearchComputer vision study
Doshisha Logo Detection
R-CNN-based logo detector using Selective Search and fine-tuned ResNet-18 for identifying the Doshisha University logo in images.
Data / MLEngineering
PythonR-CNNResNet-18Jupyter
ProductionLive · gharkatha.com
GharKatha
Smart rental management platform for Nepal — landlords and tenants manage buildings and units, generate Nepali-calendar invoices, track rent and electricity payments (eSewa / Khalti / bank), and handle maintenance requests in one place.
EngineeringInfra
Next.jsTypeScriptReactNode.jsPostgreSQL
Get In Touch
Contact
Currently based in Osaka, Japan. I am open to PhD opportunities, research collaborations across Japan, and discussions about AI, NLP, data science, and healthcare. Feel free to reach out.