About
Software Developer & AI Engineer
Hi, I'm Binaya Sharma, a software developer with strong experience in backend systems, big data, and AI-driven applications. I hold a Master's in Computer Science from LSU and have worked on projects ranging from LLM deployment and evaluation to large-scale microservices and data pipelines. I focus on building robust, efficient systems that solve real problems. Outside of tech, I am an avid traveler who enjoys exploring new places and cultures, bringing a global perspective to my work.
Skills
My work spans scalable backend systems, distributed data pipelines, and applied AI, with hands-on experience building and deploying production-grade software and domain-specific LLMs.
Backend & Systems
Databases
AI & Large Language Models
Big Data & Distributed Systems
Cloud & DevOps
Engineering Practices
Resume
I have worked on projects that translate research and engineering into real-world impact, including leading the development of production-grade reporting systems for law enforcement at 365 Labs, building large-scale big data pipelines for the European Central Bank (ECB), developing microservices for pharmaceutical R&D, designing a domain-specific language model for digital forensics, and applying advanced machine learning techniques to predict vehicle energy consumption across urban road networks.
Education
Master's in Computer Science
2023 - 2024
Louisiana State University, Baton Rouge, LA
Relevant Courses: Artificial Intelligence, Algorithms Analysis and Design, Cloud Computing, Big Data, Advanced Operating Systems
Bachelor's in Computer Engineering
2015 - 2019
Tribhuvan University, Kathmandu, Nepal
Relevant Courses: Data Structures and Algorithms, Theory of Computation, C/C++, Database Systems, Enterprise Applications, Project Management
Current Position
Senior Software Developer
2025 - Present
365 Labs, Baton Rouge, LA
- Led end-to-end development of Incident-Based Reporting (IBR) module using C#, .NET, and Entity Framework, delivering implementations across 9 states within 10 months for law enforcement RMS systems.
- Translated 200+ pages of state-specific technical documentation into functional business logic and backend architecture, breaking down complex requirements into sprint-ready tasks for Agile development cycles.
- Collaborated with cross-functional teams including product managers, QA engineers, and UI/UX designers to ensure seamless integration of IBR features into existing RMS platforms.
- Coordinated 4-member engineering team through code reviews, pull request evaluations, and sprint planning, ensuring high-quality deliverables and adherence to software development best practices throughout project lifecycle.
- Designed and maintained UI components with UWP, developed comprehensive unit tests, resolved production bugs, and provided client support to achieve high system reliability and user satisfaction.
- Automated database environment setup with custom seeding scripts, reducing manual configuration effort by 70% and eliminating environment-related bugs, streamlining deployment processes for development teams.
Past Professional Experiences
Graduate Research Assistant
2023 - 2024
LSU, Baton Rouge, LA
- Implemented Residual Vision Transformer (RVT) to predict energy consumption pattern for multiple vehicles across a city’s road network, achieving 20% improvement to Convolution Vision Transformer (CVT).
- Migrated web app and data for Artifact Genome Project from public (DigitalOcean) to private server, reducing operational costs and increasing data security.
- Developed one of the first local Large Language Model (LLM) for digital forensics by fine-tuning LLaMA-3-8B using the Retrieval Augmented Fine-tuning (RAFT) approach on a corpus of digital forensics research papers and curated digital artifacts, achieving 4.06% improvement in BERTSCORE F1, 15.79% improvement in G-Eval (using GPT-4o) and 86.6% source citation accuracy.
- Conducted user survey to evaluate practical applicability of ForensicLLM by developing a survey web app using Flask and deployed on Azure.
- Deployed multiple model configurations (base, fine-tuned, RAG) and managed concurrency using vllm to prevent memory crashes on an NVIDIA RTX 4090 handling over 80 participants.
- Performed Kruskal-Wallis tests to assess statistical significance of the user ratings across categories such as participant’s digital forensics experience and familiarity with LLMs.
Big Data Developer
2022 - 2023
Ultra Tendency, Kathmandu, Nepal
- Collaborated with international teams remotely to refactor big data pipeline for European Central Bank (ECB) using Cloudera Distributed Hadoop platform (Kafka, HBase, Spark) and Spring Boot.
- Improved code coverage by over 20% by resolving critical bugs and implementing test cases.
- Managed microservices deployments with Ansible and Kubernetes.
Software Developer
2019 - 2021
Leapfrog Technology Inc, Kathmandu, Nepal
- Developed microservices for pharmaceutical R&D utilizing AWS cloud infrastructure (EC2, RDS, S3).
- Developed RESTful APIs using Micronaut and Hibernate, with Oracle backend and lit-html frontend.
- Integrated and optimized BLAST search algorithm using multi-processing, reducing comparison time by 50%.
- Implemented secure authentication using JWT and dynamic role-based authorization across microservices.
- Automated data migration processes from multiple sources streamlining transitions and minimizing manual labor.
- Contributed to API’s documentation of different microservices used using Swagger API tools.
business Intelligence (bi) intern
2019 - 2019
LIS Nepal, Lalitpur, Nepal
- Understand the ETL process being used for a chain retail company.
- Write Python and SQL scripts to carry out ETL processes.
- Generate BI reports using MicroStrategy.
Portfolio
- All
- Project
- Research Paper
ForensicLLM
A local Large Language Model (LLM) for digital forensics developed by fine-tuning LLaMA-3-8B using the Retrieval Augmented Fine-tuning (RAFT) approach on a corpus of digital forensics research papers and curated digital artifacts, achieving 6.9% improvement in BERTSCORE F1 and enabling the model to accurately cite sources 84.6% of the time. 2024
DeepTran
DeepTran leverages deep learning techniques to accurately forecast energy consumption for various vehicle types across a city's road network, incorporating real-time traffic conditions. It integrates a Residual Neural Network (ResNet) into the Vision Transformer architecture to capture spatial patterns and temporal dependencies. 2023
Backward Pattern Matching using FM-Index
The FM index is a specialized index that combines the Burrows-Wheeler Transform (BWT) with a compressed data structure, making it useful for locating specific patterns and determining the frequency of pattern occurrences in large string sequences. This indexing strategy is very effective when searching for a specific pattern, such as locating a certain gene in a DNA sequence. 2023