Portfolio — Urlabari, Nepal
Full-stack MERN engineer → AI engineer in training. Production web systems, shipped end to end. Now crossing from writing every rule to training the model — a full-stack engineer becoming an AI engineer.
Scroll — the journey begins
Ch.01 — Foundation
Years inside the MERN stack — designing schemas, securing APIs, and shipping React frontends that hold up in production. Real clients, real deadlines, no safety net. This is the bedrock the next chapter stands on.
Full-Stack Engineer → AI Engineer
Dipak Sharma
Capabilities
Experience & education
Freelance & client projects
End-to-end delivery for local businesses and startups — React/Next.js frontends, Node/Express backends, deployment and ongoing maintenance.
Tribhuvan University · Model Campus Damak
Software engineering foundations, data structures, database systems and web technologies.
Nepal Darshan College
Completed with a 3.52 GPA.
E-commerce platform for a furniture brand — catalogue, filtering, and inquiries. Structured product data so listings index cleanly, and a backend the owner can operate without a developer.
“The e-commerce platform improved our online inquiries and gave us a professional digital brand presence.”
Founder, Sharma Furnitures
Restaurant platform frontend — menu and order management built for fast, repetitive daily use. Typed component architecture so the surface stays maintainable as features land.
Signals
“Dipak delivered a high-performance news platform with clean architecture and reliable deployment workflows.”
Product Manager
Guhaar Media
“The e-commerce platform improved our online inquiries and gave us a professional digital brand presence.”
Founder
Sharma Furnitures
“Excellent communication, strong ownership, and consistent shipping speed across both frontend and backend.”
Team Lead
Scripcol
Ch.03 — The Shift
Deterministic
The turn
Probabilistic
Ch.04 — The Frontier
The transition is underway — a deliberate curriculum from machine-learning fundamentals to production AI systems, built on an engineer's instinct for shipping.
Python & ML foundations
NumPy, pandas, and the math under the models.
Deep learning · PyTorch
Networks built — and broken — by hand.
LLM engineering
Prompting, tool use, and agent patterns.
RAG & vector search
Retrieval pipelines over real data.
Fine-tuning & evals
Adapting models, measuring honestly.
MLOps & serving
From notebook to production endpoint.
The MERN years carry over: APIs become model serving, databases become vector stores, and product instinct doesn't retrain from scratch.
Final chapter — Open channel
Roles, collaborations, or a conversation about the road from software to AI — the channel is open.
Open a channel