About Me
🔠I’m currently working on Speaker Diarization.
🌱 I’m currently learning about MongoDB, Graph VQA, Docker and Kafka.
📫 You can reach me via email at spalkhiw@asu.edu.
📄 Know about my experiences - Resume
âš¡ Fun fact: I hold a Bachelor’s degree in Electronics and Communication
Work Experience
Machine Learning Intern @ Mirwork Inc | (June 2024 - Present)
- Engineered and optimized data extraction pipelines using YouTube API to process audio, gathering about 70+ hours of content.
- Designed, implemented, and maintained MongoDB database to efficiently store and manage extensive audio data sets.
- Utilized YouTube data to fine-tune LLM to improve the accuracy and effectiveness of interview assessment tools.
- Implemented Speaker Diarization techniques for structuring data, enhancing LLM fine-tuning for interview assessments.
AI Integration Specialist @ Psych for Life | (August 2023 – Present)
- Achieved 80% reduction in research time by utilizing Prompt Engineering for enhanced efficiency.
- Leading integration of AI into the writing process for enhanced efficiency, resulting in a streamlined timeframe.
- Orchestrated the development and implementation of custom prompt engineering solutions, resulting in a 50% reduction in process cycle time and a 30% increase in team productivity.
- Conducted research, literature reviews, data analysis, and fact-checking for 65+ scientific documents.
- Spearheaded collaboration efforts between internal teams and Luminosity Labs for development of a scalable platform, streamlining operations and improving cross-departmental communication.
- Employed caching techniques, resulting in reduction in 70% decrease in rendering delay and improving website performance
Computer Vision Intern @ eInfochips (An ARROW Company) | (Jan 2023 – June 2023)
- Led a team of 3 to develop a real-time vehicle detection system for visually impaired individuals, achieving 65.4% average precision and 70.7% for car detection using YOLOv5.
- Achieved an 86.5% reduction in resource usage by filter pruning algorithm while maintaining a mean precision of 42%.
- Computed CNN models on 16,000 image datasets for vehicle detection, designed to aid visually impaired individuals.
- Enhanced real-time safety detection in YOLO by implementing Safety Detection Algorithm, achieving a 25 fps detection rate.
- Applied precision-focused strategies and fine-tuned the model using Indian traffic data to enhance accuracy in detecting vehicles within specific environmental conditions.
- Collaborated with engineers to streamline training processes, document methodologies, and advance computer vision technology for traffic surveillance and safety applications.
Software Engineering Intern @ Oxvi Respire Solutions | (May 2022 – July 2022)
- Developed an Android app to plot real-time values from a prototype ventilator with live Data Visualization.
- Integrated Firebase for real-time communication, enabling data transmission between the device and application.
- Implemented communication protocols to embedded C components in the prototype for data exchange within a 5ms lag.
Projects
Graph-based QA or Intelligent Traffic Analysis | (July 2022 - Dec 2022)
- Extracted 70,000+ frames from traffic videos at 10 fps, enhancing usability and improving VQA system accuracy
- Converted 10,000+ questions into GloVe word embeddings, improving the model’s NLP and question understanding capabilities for Scene Graph Generation
Hate Speech Detection using NLP and AI | (Jan 2024 - Present)
- Managed a team of 9 in developing and implementing NLP and AI-driven solutions for content filtering.
- Delegated research by analyzing over 50+ review articles and research papers to enhance content moderation.
Comparative Analysis of ML Models for Student Grade Prediction | (July 2022 - Dec 2022) Publication
- Analyzed and remodeled multiple classifiers, Random Forest and Decision Tree, by optimizing hyperparameters to improve accuracy by 6% and F1-score by 12% for predicting students’ grades.
- Demonstrated understanding of Machine Learning fundamentals, including data preparation, model selection, and performance evaluation, with model selection achieving 87.3% accuracy and 0.913 F1-score.
- Enhanced hyperparameters tuning using GridSearchCV, leading to a significant 41% improvement in F1 score.
- Conducted comprehensive analysis of student grades to identify correlations between different factors and academic performance, aiming to assist educational institutions in improving student outcomes.
Fault Prediction for Combination Circuits | (June 2022 - Nov 2022)
- Evaluated and deployed Machine Learning models like Support Vector Classifier, KNN for fault prediction in circuits.
- Accelerated testing processes and slashed time by 25% spent on fault testing and verification.
- Accomplished 36% accuracy while predicting faults, which was improved to 54% by Data Engineering.
- Optimized performance by Feature Engineering, leading to a 40% reduction in computation time.
Movie Booking System | (Sept 2021 - Nov 2021)
- Collaborated on a C program utilizing Arrays and Queues to simulate theater seating arrangement, allowing users to choose seats and rows and display the total amount due.
Education
- M.S., Computer Science - Arizona State University (May 2025)
- B.Tech., Electronics and Communication - Nirma Institute of Technology (May 2023)
Technical Skills
- Languages : C++, Python, C, Perl, Shell, Kotlin, Linux, Java, Matlab
- Machine Learning Techniques: Regression, Statistics, Gradient Boosting, Deep Learning, Hyperparameter Tuning, Feature Engineering.
- Tools & Frameworks : HTML, CSS, YOLO, OpenCV, Flutter, Kafka, Apache, Spark, Docker, SQL
- Developer Tools : AWS, Git, Anaconda, Spyder, Linux, IntelliJ, VSCode, Android Studio, LaTeX, PowerBI, Tableau, Firebase, Whisper
- Libraries : NumPy, Pandas, Keras, Scikit-learn, Torch, TensorFlow, PyTorch, Neural Networks, Pyannotate, Seaborn
Publications
- Analysis of Machine Learning algorithms for predicting student grade - Springer Journal of Data, Information and Management
- Fire Detection and Control Systems - STM Journals, International Journal of Radio Frequency Design