Shishir Khanal

Summary:

An interdisciplinary engineer with a solid foundation in engineering, artificial intelligence (AI), and data science, backed by a diverse academic background across the USA, UK, India, and Nepal. Experienced in engineering, research, AI-driven projects, financial modeling, mechanical design, and system development, coupled with finely tuned technical writing, critical thinking, and problem-solving skills. Proficient in AI/ML, deep learning, NLP, generative AI, large language models (LLMs), programming languages, MS Office, CAD (3D & 2D), and mechanical engineering.

Education:

Master’s Degree in Engineering, USA
- Mathematical and Computational Modeling • Data Visualization and Analysis • Stochastic Modeling • Derivative Pricing • Python Programming • Machine Learning • Deep Learning • Portfolio Management • Organization and project management • Risk Management • Marketing, Finance, Accounting, and Human resources (HR) • Master Thesis
Master’s Degree in Artificial Intelligence (AI), UK
- Statistics and EDA • Machine Learning • Deep Learning • Computer Vision • Natural Language Processing (NLP) • Reinforcement Learning • Generative AI • LLM • Master Thesis
Executive Post Graduate in Machine Learning & Artificial Intelligence, India
- Python Programming • Mathematics for AI/ML • Statistics and Exploratory Data Analytics • Data Analysis using SQL and Excel • Data Visualisation • Machine Learning • Deep Learning • Computer Vision • Natural Language Processing (NLP) • Reinforcement Learning • Generative AI • LLM • Capstone
Bachelor’s Degree in Mechanical Engineering, Nepal
- Engineering Mathematics • Statistics and Probability • Computer Programming • Control Systems • Numerical Method • CAD • Engineering Drawing • Project Engineering and Management • Engineering Professional Ethics • Electrical and Electronics Engineering • Metrology • SOM • Instrumentation and Measurement • Applied Mechanics • Thermodynamics • Fluid Mechanics • Theory of Machines and Mechanisms • Machine Design • Heat and Mass Transfer • Manufacturing Engineering • Material Science • Internal Combustion Engines • Renewable Energy Systems • Automobile Engineering • Hydraulics and Pneumatics • Industrial Engineering • Engineering Physics • Engineering Chemistry • Engineering Economics • Kinematics of Machines • Refrigeration and Air Conditioning • Finite Element Method • Project Work

Work Experience:

System (AI) Development and Design Engineer at Tokyo, Japan-(2020 to present)
Mechanical Engineer at Tata Motors-(2018 to 2019)
Internship Engineer at Pepsi-2017

Machine Learning:
- Machine learning algorithms and techniques, including Regression (Linear, Logistic, Polynomial, Ridge and Lasso Regression), Classification (Decision Trees, Random Forest, Support Vector Machines, Naive Bayes, K-Nearest Neighbors, Gradient Boosting), Clustering (K-Means, Hierarchical Clustering, DBSCAN, Gaussian Mixture Models)
- Model evaluation techniques (Cross-validation (K-Fold, Stratified K-Fold), Confusion Matrix, Precision, Recall, F1 Score, ROC-AUC, Precision-Recall curves, Mean Absolute Error, Mean Squared Error, RMSE, R² Score) and predictive modeling
- Hyperparameter tuning and optimization techniques (Grid Search, Random Search, Bayesian Optimization, Genetic Algorithms)
- Reinforcement learning and mathematical modeling
- Ensemble methods (Boosting, Bagging, AdaBoost, Gradient Boosting)
- Dimensionality reduction techniques (Principal Component Analysis, t-SNE)

Deep Learning:
- Neural Networks, Feedforward, Convolutional Neural Networks (CNN), Recurrent Neural Networks(RNN), Long Short-Term Memory, Gated Recurrent Units (GRU)), Autoencoders, Transformers, Attention Mechanisms, Transfer Learning, Generative Adversarial Networks
- Optimization Techniques (Backpropagation, Stochastic Gradient Descent (SGD) and variants (Adam, RMSprop, AdaGrad), Batch Normalization, Dropout, Layer Normalization)
- VGG, ResNet, Inception, EfficientNet
- Computer Vision, Image Processing, Object Detection, Segmentation
Natural Language Processing (NLP):
- Lexical analysis, Syntactic analysis, Semantic analysis, Discourse integration, Pragmatic analysis
- Text Preprocessing: Tokenization, Lemmatization and Stemming, Stopword removal, Sentence splitting, Part-of-Speech (POS) tagging, Named Entity Recognition (NER), Text normalization (lowercasing, punctuation removal)
- Text Representation: Bag of Words (BoW), Term Frequency-Inverse Document Frequency (TF-IDF), Word embeddings (Word2Vec, GloVe, FastText), Inside-outside-beginning (IOB), Sentence embeddings (BERT, GPT, Sentence-BERT, etc.), One-hot encoding
- Language Models: Transformer models (BERT, GPT, RoBERTa, T5), Recurrent Neural Networks (RNN, LSTM, GRU), Attention mechanisms, Pre-training and fine-tuning of large models
- Text classification, Sentiment analysis, Topic modeling, Emotion detection, Sequence modeling, Machine translation, Text generation, Question answering, Text summarization, Semantic role labeling, Co-reference resolution, Dependency parsing, Intent detection, Slot filling, Response generation, Content creation
- Text Similarity & Clustering: Cosine similarity, Jaccard similarity, Text clustering (e.g., K-means, Agglomerative clustering), Topic modeling (LDA)
- Evaluation metrics: precision, recall, F1 score, BLEU, ROUGE, METEOR, perplexity, cross-entropy
- Frameworks & Libraries: TensorFlow, PyTorch, Hugging Face Transformers, spaCy, NLTK, Gensim, OpenAI API, Stanford NLP, speech recognition (ASR models), text-to-speech (Tacotron, WaveNet)

Generative AI:
- Generative Models: GANs (Generative Adversarial Networks), VAEs (Variational Autoencoders), Diffusion Models, DCGAN (Deep Convolutional GAN), ProGAN (Progressive GAN), StyleGAN, Pix2Pix, Autoencoders, Seq2seq Models, WaveNet, Text-to-Image Models (like DALL·E), DDPM (Denoising Diffusion Probabilistic Models)
- Transformer-based Models: Transformers, GPT (Generative Pretrained Transformer), BERT (Bidirectional Encoder Representations from Transformers), T5 (Text-to-Text Transfer Transformer), LLMs (Large Language Models), SLMs (Smaller Language Models), Attention Mechanism (Transformer Architecture)
- RNN-based and Sequential Models: LSTM (Long Short-Term Memory), RNNs (Recurrent Neural Networks), Encoder-Decoder Models
- Model Optimization and Techniques: Fine-tuning, Prompt Engineering, LangChain, LangChain AI Agent, LlamaIndex (for efficient retrieval in generative tasks) RAG (Retrieval-Augmented Generation)
- Information Retrieval and Extraction: Named entity recognition (NER), Relation extraction, Document retrieval (using vector databases and embeddings), Knowledge graph creation, Retrieval-Augmented Generation (RAG)
- Conversational AI, Fine-tuning and Transfer Learning, Data Annotation and Management, Evaluation Metrics, Ethics in AI, Diffusion Models, AI Agent, Multimodal Models, Document Question Answering, Imagen
- OpenAI APIs, ChatGPT, BERT, Gemini, and Hugging Face Transformers

Data Handling & Analysis:
- Exploratory data analysis (EDA), Data preprocessing, feature engineering, and data visualization and analysis
- SQL for data querying and management
- Kaggle for data science competitions and projects
Mathematics and Statistics:
- Linear Algebra, Statistics, Probability, Calculus, Differential Equations, Numerical Methods, Geometry, Trigonometry, Vector Calculus, Complex Numbers, Mathematical Modeling.

$Image 1$

Software Development & Collaboration:
- Version control using Git and GitHub
- Web technologies: HTML, CSS, JavaScript
- Microsoft Word, Excel, and PowerPoint for documentation and presentations
Programming & Libraries:
- Data manipulation: Numpy, Pandas
- Data visualization: Matplotlib, Seaborn, Plotly, Bokeh
- Machine learning frameworks: Scikit-learn, TensorFlow, Keras, PyTorch, XGBoost, LightGBM
- Computer vision: OpenCV
- NLP: spaCy, NLTK, Hugging Face Transformers

Mechanical Engineering:
- Mechanical Design & Product Development, proficiency in SolidWorks, CATIA, ANSYS, and AutoCAD (2D & 3D)
- Structural Analysis & FEA (Stress Analysis, Fatigue, Thermal Analysis)
- Thermodynamics & Heat Transfer (HVAC, Heat Exchangers, Cooling Systems)
- Fluid Mechanics & Hydraulics (CFD, Pumps, Compressors, Pipeline Systems)
- Manufacturing & Production Engineering (CNC Machining, Casting, Welding, Additive Manufacturing)
- Materials Science & Metallurgy (Material Selection, Failure Analysis, Surface Treatments)
- Pneumatic & Hydraulic Systems (Braking, Steering, Actuators, Fluid Power)
- Kinematics & Dynamics of Machines (Vibration Analysis, Mechanism Design, Motion Simulation)
- Quality Assurance & Control (ISO Standards, Six Sigma, Lean Manufacturing)
- Metrology, Instrumentation, and Measurement Techniques
- Mechatronics & Control Systems (Sensors, Actuators, Embedded Systems)
- Project Management & Technical Documentation (DFMEA, Risk Assessment, Compliance)

Automotive Engineering & Vehicle Technologies:
- Automotive Design & Powertrain Engineering (IC Engines, Transmissions, EV Powertrains)
- Diagnostics, Repair & Maintenance (Fault Detection, Troubleshooting, Preventive Maintenance)
- Chassis, Suspension & Drivetrain (Multi-Link, Air Suspension, Adaptive Damping)
- Tire Dynamics & Wheel Alignment (TPMS, Wear Analysis, Performance Optimization)
- Battery & High-Voltage EV Systems (Charging, BMS, Thermal Management)
- Electrical & Electronics Systems (ECUs, CAN Bus, ADAS, Wiring Harness)
- Cooling & HVAC Systems (Engine Cooling, Climate Control, Heat Exchangers)
- Hydraulic & Pneumatic Systems (Brake Systems, Power Steering, Air Suspension)
- Vehicle Dynamics & Performance Analysis (Aerodynamics, Handling, Stability)
- NVH (Noise, Vibration & Harshness) Optimization (Soundproofing, Vibration Control)
- Heavy Vehicle Systems & Compliance (Emissions Control, DPF, SCR, AdBlue, Regulatory Standards)
- Prototyping, Testing & Validation (Durability Testing, Crash Testing, Regulatory Compliance)
- Manufacturing & Production Processes (Assembly, Material Selection, Quality Control)
- Automotive Safety & Risk Assessment (Crashworthiness, Safety Protocols, ISO Compliance)

Soft Skills:

Problem-Solving & Critical Thinking (Root Cause Analysis, Innovative Solutions)
Communication & Interpersonal Skills (Team Collaboration, Client Interaction, Technical Presentations)
Leadership & Team Management (Project Leadership, Mentoring, Conflict Resolution)
Creativity & Innovation (Concept Development, Process Improvement, Design Thinking)
Decision-Making & Risk Management (Evaluating Trade-offs, Analyzing Potential Outcomes)
Collaboration & Cross-Functional Teamwork (Working with Diverse Teams, Cross-Departmental Coordination)
Time Management & Organization (Task Prioritization, Deadline Management, Multitasking)

Adaptability & Flexibility (Adjusting to New Technologies, Working in Dynamic Environments)
Attention to Detail & Accuracy (Precision in Design, Error Detection, Quality Assurance)
Self-motivation (Dedication, responsibility, and reliability in completing tasks)
Cultural Sensitivity (Respecting Diverse Backgrounds, Adapting to Global Teams, Promoting Inclusivity)
Leadership & Management (Leading Teams, Motivating Others, Providing Clear Direction)
Empathy & Emotional Intelligence (Understanding Team Needs, Building Rapport)

Customer Focus & Client Relationship Management (Customer-Centered Solutions, Active Listening)
Conflict Resolution & Negotiation (Mediating Disagreements, Reaching Consensus)
Stress Management (Staying Calm Under Pressure, Managing Stressful Situations, Prioritizing Tasks)
Customer Service (Understanding Client Needs, Providing Solutions, Maintaining Positive Relationships)
Positive Attitude (Maintaining Optimism, Encouraging Team Morale, Fostering a Can-Do Mindset)
Confidence (Speaking Assertively, Taking Initiative, Trusting One’s Abilities)
Continuous Learning & Self-Improvement (Staying Updated on Industry Trends, Pursuing Skill Enhancement)

Languages skills:

English: Working Proficiency Plus (TOEIC Certified)
Japanese: Business Level Proficiency (JLPT-N2 Certified)
Hindi: Intermediate Level Proficiency
Nepali: Native or bilingual proficiency

Certificates:

Certificate for Registered Engineers
AI Qualification Certificate by the Japan Deep Learning Association (JDLA)
Japan Ministry of Economy, Trade and Industry Observer - DX Promotion Passport1 (AI) by “Di-Lite” Japan
Generative AI Passport by Generative AI Utilization Promotion Association (GUGA), Japan
AI Literacy Exam: Ranked in the Top 1%, GUGA, Japan
The Deep Learning Specialization, DeepLearning.AI
Crash Course on Python, Google
Linear Algebra for Machine Learning and Data Science, DeepLearning.AI
Neural Networks and Deep Learning, DeepLearning.AI
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization, DeepLearning.AI
Convolutional Neural Networks, DeepLearning.AI
Sequence Models, DeepLearning.AI
Introduction to Artificial Intelligence (AI), IBM
Programming for Everybody (Getting Started with Python), University of Michigan
The Smithsonian Entrepreneur Incubator Course, US Embassy
Engineering and Product Design Processes, Arizona State University

Mechanical Design, Solidworks:

AI Literacy Exam (Japan): Ranked in the Top 1%

AI Qualification Certificate By Japan Deep Learning Association（JDLA）

Generative AI Qualification Certificate By Japan Deep Learning Association（JDLA）

DX Qualification Certificate By Digital Literacy Council, Japan

Generative AI Passport By Generative AI Utilization Promotion Association (GUGA), Japan

OpenBadges:

Click Here for Details

Click Here for Details

ML/AI Projects:

Gen AI Project: Generative Search System (Mr.HelpMate AI):

Architecting a sophisticated Generative Search System named Mr.HelpMate AI, utilizing advanced Retrieval-Augmented Generation (RAG) techniques and state-of-the-art Hugging Face models. This project integrates cutting-edge frameworks such as LangChain and LlamaIndex to enhance the search experience through dynamic and contextually relevant responses. By combining powerful generative models with retrieval mechanisms, Mr.HelpMate AI aims to provide highly accurate and context-aware search results, enabling users to efficiently find and interact with information across diverse domains. This system leverages the latest advancements in AI to deliver intelligent and responsive search capabilities, transforming how users engage with digital content.

AI-Driven Melanoma-Detection System:

Engineering a state-of-the-art multiclass classification model utilizing a bespoke Convolutional Neural Network (CNN) architecture within TensorFlow to enable automated detection and classification of skin cancer. This project employs advanced deep learning methodologies, including sophisticated feature extraction and hierarchical neural network layers, to discern and differentiate melanoma and other dermatological malignancies with high precision. By integrating cutting-edge image preprocessing techniques and model optimization strategies, the system aims to deliver a highly reliable diagnostic tool that supports dermatologists in early and accurate cancer detection, thereby significantly enhancing clinical decision-making and patient management through automated, real-time analysis.

AI-Driven Telecom Churn Prediction System:

Designing a churn prediction model for the telecommunications industry using advanced machine learning techniques. This project focuses on developing and optimizing predictive models to accurately forecast customer attrition, leveraging a range of algorithms including Logistic Regression, Decision Trees, and Random Forests. By systematically evaluating model performance through rigorous metrics and aligning with business objectives, the system aims to identify key predictor variables that contribute to customer churn. The ultimate goal is to implement the most effective model, providing actionable insights to enhance customer retention strategies and drive business growth.

AI-Driven Gesture Recognition System:

Developing a cutting-edge gesture recognition system for smart TVs, leveraging advanced deep learning techniques to enable intuitive, hands-free control. This project involves designing and implementing a robust deep learning model to accurately identify and classify five distinct user gestures from real-time video feeds. Utilizing sophisticated neural network architectures, including convolutional layers for feature extraction and recurrent layers for temporal pattern recognition, the system aims to translate user gestures into TV commands, eliminating the need for a remote control. This innovative approach enhances user experience by providing seamless and interactive control, advancing the usability and functionality of smart home devices.

AI-Driven Automatic Classification System Of Customer Complaints:

Developing an intelligent classification system for automating the categorization of customer complaint tickets. This project leverages advanced machine learning techniques to build a solution that accurately identifies and classifies the type of complaint raised by customers from unstructured text data. By employing sophisticated natural language processing (NLP) methods and machine learning algorithms, the system aims to streamline complaint management processes, enhance operational efficiency, and improve response accuracy. The ultimate objective is to enable businesses to swiftly address customer issues and drive better customer satisfaction through automated, precise complaint categorization.

AI-Driven Stock Price Prediction System:

Developed an AI-based stock price prediction system leveraging Long Short-Term Memory (LSTM) neural networks to model and forecast future stock prices based on historical time series data. The system was designed to capture long-term dependencies and trends in financial data using techniques such as data normalization, sliding window for sequence generation, and feature engineering including moving averages and technical indicators. LSTM layers were optimized with dropout regularization and Adam optimizer to prevent overfitting and improve convergence. The model was evaluated using metrics like Mean Squared Error (MSE).

AI-Driven Telecom Churn Prediction System:

Developed a multiple linear regression model to forecast the demand for shared bikes using a dataset provided by BoomBikes, a leading bike-sharing service in the US. The project aimed to identify and analyze key factors influencing bike demand in various regions and weather conditions. The model was trained on historical data, incorporating variables such as temperature, humidity, wind speed, and seasonal effects to predict demand patterns. Feature engineering techniques, such as encoding categorical variables and scaling continuous variables, were used to improve model accuracy. Evaluated the model’s performance using metrics like R-squared and Mean Absolute Error (MAE), achieving high predictive accuracy. The insights derived can help optimize resource allocation, improve customer satisfaction, and guide future business strategies.

AI-Driven House Price Prediction System:

Designed and implemented a predictive model for house prices using advanced regression techniques, including Ridge and Lasso regression. This system leverages historical housing market data, with features such as property size, location, number of rooms, and neighborhood demographics. Ridge regression was used to mitigate multicollinearity, while Lasso regression helped with feature selection by reducing the weight of less important variables to zero. The model was optimized using cross-validation techniques and evaluated through performance metrics like R-squared and Root Mean Squared Error (RMSE), providing accurate and interpretable predictions. The system offers valuable insights for real estate investors, homebuyers, and market analysts to make data-driven decisions.

AI-Driven Fake News Detection System:

Developing a comprehensive Fake News Detection System employing a diverse array of machine learning models to combat misinformation. This project integrates several advanced classification algorithms, including Logistic Regression, Decision Tree Classification, Gradient Boosting Classifier, and Random Forest Classifier, to create a robust and reliable detection framework. By leveraging these diverse models, the system is designed to analyze textual data and identify misleading or false information with high accuracy. The approach incorporates feature engineering, model ensemble techniques, and rigorous evaluation metrics to ensure effective and scalable detection of fake news, contributing to a more informed and trustworthy information ecosystem.

AI-Driven Intelligent Symptom Checker:

Developed an AI-driven system designed to analyze user-reported symptoms and provide potential diagnoses in real-time. Utilizing advanced NLP and ML models, the system identifies and classifies symptoms to suggest possible medical conditions, as well as recommend potential treatment options. This empowers users with reliable health information while also assisting doctors and researchers in identifying patterns, enhancing early symptom evaluation, and improving healthcare accessibility. The tool supports healthcare professionals by streamlining diagnosis and treatment decisions, ultimately contributing to more efficient patient care.

AI-Driven Handwriting Detection System:

Developed an AI-powered handwriting detection system using a convolutional neural network (CNN) trained on a large dataset of handwritten images. The system efficiently classifies and recognizes digits and characters, utilizing advanced preprocessing techniques like normalization and data augmentation to enhance accuracy. With a well-optimized CNN architecture and regularization methods, the model achieved high precision, making it suitable for real-time applications.

AI-Driven Credit Card Fraud Detection System:

Developed an AI-based credit card fraud detection system using machine learning algorithms to identify fraudulent transactions. The system was trained on a highly imbalanced dataset and utilized techniques such as oversampling and feature scaling to improve detection accuracy. Key models like Logistic Regression, Random Forest, and Gradient Boosting were implemented, with hyperparameter tuning and cross-validation to enhance performance. The final model achieved high precision and recall, providing real-time fraud detection capabilities.

Developed an AI-powered assistive technology, “Eye for the Blind,” designed to help visually impaired individuals by recognizing and describing their surroundings in real-time. Leveraged deep learning techniques, including object detection models like YOLOv5, and natural language processing (NLP) for converting visual data into spoken descriptions. The system integrates a camera to capture images, processes them with the trained model, and provides auditory feedback via speech synthesis.

AI-Driven Heart Failure Prediction System:

Developed a machine learning model to predict heart failure based on key clinical parameters, including anaemia, creatinine phosphokinase levels, diabetes, ejection fraction, blood pressure, platelets, serum creatinine, and serum sodium. This AI-driven system analyzes these indicators to predict heart failure risk, aiding in early diagnosis and improving treatment outcomes. Designed to assist doctors, researchers, and healthcare professionals in making more informed decisions, it is a cutting-edge solution at the intersection of AI and healthcare.

Mechanical Research Projects:

Design and Development of Magnetic Bicycle:

The conversion of magnetic power into mechanical energy is the prime subject of study for the researcher. The project on magnetic powered hybrid bicycle is focused on the investigation of the feasibility of the mechanical energy from the magnet which use its natural properties of attraction and repulsion of the magnet poles to create rotation of the shaft which can be harnessed to do a useful work.

Design:

The strong Neodymium magnets (NdFeB40MGOe) arranged in proper inclination in flywheel and cover design and assembled in bicycle sprocket shaft produce magnetic force to assist the human effort. This investigation project was proposed by performing paper research, experiments, hardware prototype development and software simulation.

Procedure:

The flywheel design was performed in Solidworks, magnetic simulation was performed in ANSYS software (FEMM). The repulsive force between magnets in outer circumference of flywheel and inner surface of the cover provides torque in sprocket shaft in addition to manual torque. The magnetic field, resultant force on magnetic flywheel was determined for rectangular magnets in circular position by magnetic simulation. The performance characteristic of the bicycle was determined by conducting practical experiment. The final design of the prototype was selected based on the optimum result obtained from the magnetic simulation in FEMM software. The alternative magnetic shielding material was used instead of Mu metal due to unavailability in local market.

Conclusion:

The performance of the bicycle was found to be increased significantly after implementation of the magnetic flywheel. Thus, the detail study on this project has led to the need of more research and can be developed as one of means of green energy.