Baburam Adhikari
Software DeveloperData & AI
I build modern web applications and data-driven solutions using Python, JavaScript, React, SQL, and AI/ML technologies.
Edmonton, AB, Canada
from django.http import JsonResponse
import pandas as pd
def insights(request):
df = pd.read_sql("SELECT * FROM events", db)
kpi = df.groupby("segment").rate.mean()
return JsonResponse(kpi.to_dict())Conversion by segment
+18.4%
About
I'm a full-stack software developer with 2+ years of professional experience building production web applications in Python (Django, Flask) and TypeScript (React, Next.js). I've worked on Scrum teams at two software companies, owning features from the PostgreSQL schema and REST API through to the user interface.
Day to day that means authentication and authorization, database design, query tuning, testing, and code review — and taking part in the architecture conversations about how a system should be split and how it will hold up as data grows. I also bring a machine learning and computer vision background, which I've used to integrate ML models into working web applications rather than leaving them in notebooks. Two of those research projects are peer-reviewed and published.
From PostgreSQL schema and REST API through to the interface — requirements to production deployment, on a Scrum team.
API contracts the frontend develops against, code review, test coverage, and documentation — plus mentoring juniors on API design.
Computer vision and NLP models integrated into working web applications, kept in their own modules behind clean service interfaces.
What I Do
I build full-stack applications, work with databases and APIs, and integrate data analytics and AI/ML to solve real-world problems.
Building scalable, responsive, and database-driven applications.
Transforming data into insights through analysis, visualization, SQL, and automation.
Developing intelligent applications using machine learning, NLP, computer vision, and recommendation systems.
Featured Projects
Each project opens into a full case study — problem, architecture, implementation, data, results, and what I'd change next. Two are peer-reviewed and published.
A full-stack credential vault with encrypted storage, TOTP two-factor auth, and optional biometric login.
Key result
AES-256 + TOTP— Encrypted vault with layered access control
A learning platform where students run Python directly in the browser.
Key result
In-browser Python— Pyodide execution, no backend runner
A Flask web application that runs YOLOv8 detection over drone footage.
Key result
Notebook → browser— YOLOv8 inference served behind a REST API
A Pandas and SQL pipeline over employment data, published as Power BI dashboards.
Key result
10K postings— Cleaned to zero nulls, shipped as dashboards
Exploratory analysis and detector benchmarking over a large-scale banking transaction dataset.
Key result
3 detectors— Benchmarked on precision, recall, and F1
A modular recommendation backend with a trust-aware Learning-to-Rank ensemble.
Key result
LTR ensemble— SBERT + ALS + LightGBM behind a service interface
A hybrid Transformer + CNN + BiLSTM model for Devanagari and Romanized Nepali.
Key result
73.88% accuracy— Benchmarked against XLM-RoBERTa and NepaliBERT
Four-class retinal classification at 91.14% precision on a small compute budget.
Key result
91.14% precision— Across 4 retinal classes
How I Build
The projects above share a shape. Here is what I own at each layer of the stack.
Frontend
What the user touches
API
The contract between the two halves
Backend
Where the rules live
Database
The source of truth
Frontend → API → Backend → Database
Data in Practice
The analytical half of the work, drawn from the projects above. Analysis is only finished when someone can act on it, so each one runs the same spine — question, analysis, insight, recommendation.
Demand & benchmark analysis
Risk & segmentation analysis
Experience
Two and a half years building production web applications on Scrum teams. Since 2025 my focus has been the M.Sc. at Algoma University and the projects and published research above.
APAR Nepal
Backend-leaning full-stack work on a Scrum team — the REST API layer, the PostgreSQL design behind it, and the architecture and review conversations around both.
Leapfrog Technology
Frontend feature work in the Next.js and TypeScript stack, with a focus on reusable components, API integration, and keeping the codebase maintainable.
Skills
Grouped by what each set of tools is for. No progress bars — a percentage next to a language name tells a reviewer nothing.
The core of what I do — building and shipping production applications.
Schema design and query work behind every application I build.
How the work gets built, reviewed, and kept safe.
Turning raw data into something a stakeholder can act on.
Applied modeling, built into working software.
Education & Credentials
Algoma University
Graduate study alongside the research that became two peer-reviewed publications — a trust-aware learning-to-rank recommendation system and a lightweight CNN for retinal image classification.
Relevant areas
IBM Generative AI: Prompt Engineering
IBM · 2025
Microsoft Azure Machine Learning Fundamentals
Microsoft
Career Essentials in System Administration
Microsoft
A modular recommendation backend whose ranking runs as a Learning-to-Rank ensemble over SBERT, ALS, and LightGBM, behind clean service interfaces a frontend can consume.
View publication (DOI)A four-class retinal fundus classifier reaching 91.14% precision with a smaller compute footprint than standard models.
Read the paper (PDF)Resume
I'm open to opportunities in software development, data-driven applications, and AI-focused software engineering.
Open to
Contact
A role, a project, or a question about any of the work above — send a message and I'll reply within a day.