More than meets the pixel.
COMPUTER VISION · GENERATIVE AIReconstructing image detail with Super Resolution GANs.
64 × 64 → 256 × 256. A 4× reconstruction pipeline, demonstrated through saved visual comparisons.
Explore SRGANIntelligence.With a humanperspective.
I’m Bibek Rawal. I turn complex ideas into
thoughtful AI systems that work in the real world.
Ideas into impact.
Selected research and production work.
From language and vision to connected systems.
From intent
to execution.
FLOW · PRODUCTION AIAI-native workflows meet conversational analytics. Connected databases, Power BI dashboards, and KPI alerts that turn insights into action.
Built to act. Chat-created dashboards and database-backed KPI events connected to workflows and webhooks.
Explore FlownamasteRomanized Nepali
Transliteration
नमस्तेDevanagari
Translation
Hello.English
NEPALI ↔ ENGLISHLANGUAGE WITHOUT BARRIERS
Bridging languages.
Connecting people.
NATURAL LANGUAGE PROCESSINGA hybrid transliteration and translation system making information accessible to Romanized Nepali speakers.
39.84 BLEU on the held-out translation test split, with a separate news-retrieval stage.
mBART · BM25 · PyTorch · DjangoExplore NLTTA
HELPING THINGS GROW.
A healthier kind of intelligence.
DEEP LEARNING · AGRICULTUREIdentifying vegetation diseases through leaf imagery.
92.66% best validation accuracy across 15 leaf-image classes; field performance remains untested.
Explore disease detection
THE HUMAN BEHIND THE MODELS.Engineer by training.
Explorer by nature.
I’m an AI and machine learning engineer based in Kathmandu, Nepal. I’m drawn to problems where real-world inputs challenge a model’s assumptions: the many ways people type Romanized Nepali, or crop images that need to reflect local growing conditions.
My projects have taken me from language and vision experiments to connected AI workflows. Limited training compute, imperfect datasets, and the security boundaries around tools have taught me to pay attention to the whole system.
I care about making the work inspectable: what the data represents, what an evaluation actually proves, and what happens when an automated action fails. That’s the approach I want to bring to useful AI products.
A little more about me — résuméNatural language processingComputer visionDeep learningProduction AI
Always building.
Always becoming.
From reliable datasets to production AI.
The work behind the perspective.
Associate AI Engineer
Frost Digital VenturesBuilding production AI across agent orchestration, document intelligence, and enterprise automation.
- Flow. Architected an AI-agent-native, low-code workflow platform with MCP and integrations across Teams, Slack, WhatsApp, and Telegram.
- Document intelligence. Improved table and figure detection by approximately 30% while reducing incorrect figure segmentation in an OCR platform.
- DocPilot. Designed a multi-agent customer assistance system with retrieval grounding and confidence-based escalation, reducing human intervention by 80%.
- Automated code review. Built an agentic merge-request workflow that flags conflicts and security issues, posts actionable inline comments, and notifies developers before human approval.
More about this role at Frost Digital Ventures
- Built multi-step retrieval pipelines with query decomposition and re-ranking; fine-tuned open-weight LLMs using LoRA, QLoRA, and PEFT.
- Evaluated DETR- and YOLO-based models for document layout analysis and deployed merge-request monitoring through GitLab Runner.
- Delivered FastAPI services and Docker/Kubernetes deployments, owning AI implementation end-to-end for assigned financial services and retail projects.
MCP · Multi-agent systems · RAG · FastAPI · Docker · Kubernetes
AI Tech Intern
GaoTek Inc.Took customer-support and SEO applications from proof of concept to deployed products in under two months.
- CCRAG. Built an action-oriented customer-care chatbot with retrieval-augmented generation and tools for live actions such as KYC updates.
- Retrieval evaluation. Tested retrieval quality using Precision@7, Recall@7, MRR, and nDCG.
- AI-assisted SEO. Developed a website evaluation tool that analyzes public pages and recommends SEO improvements without requiring source-code access.
More about this role at GaoTek Inc.
- Used Playwright and Selenium to extract website content for SEO analysis and knowledge-base creation.
- Delivered the chatbot and SEO application as containerized, deployed products, from initial proof of concept through implementation.
Agentic RAG · Retrieval evaluation · Playwright · Selenium · Docker
Data Specialist
Sprouts Technology (CloudFactory)Created and refined the image, spatial, and 3D data that computer vision systems depend on.
- Geospatial data. Contributed to Nearmap’s 3D roof detection project through data annotation and 3D modeling for large-scale geospatial datasets.
- Quality and delivery. Exceeded project-defined quality and productivity targets while maintaining consistent annotation standards.
- Workflow improvement. Identified annotation challenges and internal-tool bugs, collaborating with project teams to improve workflows and dataset quality.
More about this role at Sprouts Technology (CloudFactory)
- Participated in project discussions, review sessions, and management meetings to align requirements, quality standards, and delivery timelines.
Data annotation · 3D modeling · Computer vision · Geospatial datasets
B.Sc. Computer Science & IT
Tribhuvan University · Bhaktapur Multiple CampusGraduated with Distinction · 83.64%
Ranked in the top two of the cohort. Final-year NLTTA project: 98/100.
Scholarships & continued learning
- Jehendar Scholarship — Tribhuvan University Academic Excellence Scholarship; ranked first on campus.
- Merit Scholarship for Technical Bachelor Studies — Directorate of Education Development, Karnali Province.
- Participated in academic–industry dialogue at Bhaktapur Multiple Campus and AI in Healthcare training with NAAMI and B.P. Eye Foundation.
There’s a world
beyond the screen.
New trails. Different perspectives. Space to think.
The curiosity doesn’t stop when the laptop closes.
Let’s make
something matter.
Let’s talk about AI and machine learning engineering roles, or collaborate on applied AI—from language and vision to reliable automation. Tell me about your team, the problem, and what you want to build.
bibek.rawal.588@gmail.com Based in Nepal. Thinking beyond borders.