SHISHIR RIJAL
Machine Learning Engineer
Computer Engineer building intelligent systems at the intersection of deep learning, neural architectures & production AI.
Who I Am
Engineering Intelligence
I'm a Computer Engineer turned Machine Learning Engineer with a passion for building AI systems that actually ship to production. From neural network architectures to scalable inference pipelines, I thrive at every layer of the ML stack.
My work spans computer vision, natural language processing, and generative AI — designing models that are not just accurate, but fast, robust, and production-ready.
When I'm not training models or optimizing latency, I'm exploring the frontier of LLMs, diffusion models, and whatever comes next.
Tech Arsenal
Featured Projects
NepaliVisionPlus
Engineered a dual-encoder image captioning framework combining ResNet50 and InceptionV3 to extract robust spatial features and visual embeddings from input scenes.
Designed a Transformer decoder mapping visual feature vectors to linguistic sequences, generating descriptive Nepali captions with a BLEU-1 score of 0.52 on the validation set.
Integrated a custom Tacotron2-based Text-to-Speech engine, converting generated captions into natural Nepali speech — a full multimodal accessibility pipeline.
Real-Time Object Detection at the Edge
Optimized YOLOv8 for deployment on resource-constrained edge devices using quantization, pruning, and TensorRT. Achieved 45 FPS on NVIDIA Jetson Nano with only 2% accuracy drop.
LLM Fine-Tuning Framework for Low-Resource NLP
Developed a parameter-efficient fine-tuning framework using LoRA and QLoRA for adapting large language models to domain-specific tasks with limited labeled data.
Anomaly Detection System for Industrial IoT
Built a time-series anomaly detection system using transformer-based autoencoders and contrastive learning, deployed in production monitoring manufacturing equipment.
Experience & Education
- ›Building a Korean educational auto-scoring system using LLMs and custom evaluation pipelines.
- ›Conducting research for Korean to optimize scoring reliability and linguistic analysis.
- ›Fine-tuning models and deploying ML services for research and prototyping use.
- ›Awarded RMC Mini Research Grant for 'Deep Learning Study for Early Detection of Knee Osteoarthritis from Radiographic Images'.
- ›Developed CNN and transformer-based pipelines for image processing, feature extraction, and model optimization.
- ›Conducted experimental evaluation and contributed to academic writing through to final defense.
- ›Selected for competitive AI fellowship with 5% acceptance rate, completing intensive 6-month program.
- ›Designed and optimized deep learning models in computer vision.
- ›Built an in-browser AI proctoring system integrating face recognition, real-time object detection, head pose estimation, and audio-based anomaly detection.
- ›Designed, trained, and deployed scalable machine learning models for production AI systems.
- ›Worked on data preprocessing, feature engineering, model validation, and performance.
- ›Collaborated with data engineering teams on model integration, A/B testing, and reliable end-to-end ML pipelines.
- ›Developed mobile applications implementing MVVM architecture for modular codebases serving 13K+ users.
- ›Integrated RESTful APIs with error handling and offline caching, improving performance by 13%.
- ›Final grade: 80.08% (Distinction) · Class Rank: 3rd / 48
- ›Coursework: Linear Algebra, Artificial Intelligence, Data Structures & Algorithms, Computer Networks & Security, Big Data Technologies, Data Mining, Probability and Statistics.
- ›Final grade: 3.82 / 4.0 · Class Rank: 2nd / 200+
- ›Final grade: 3.75 / 4.0 · Class Rank: 1st / 32
Verified Certifications
Let's Connect
Building something interesting?
Let's talk.
Whether it's a challenging ML problem, a research collaboration, or just a conversation about the state of AI — I'm always open to connecting with curious minds.