Joshua Jose

Entry-level Data Scientist with practical experience in real-time computer vision and Generative AI solutions through academic and industry externships. Developed and implemented YOLO-based weapon detection and FaceNet+SVM face recognition systems for prison surveillance, achieving sub-200 ms alerting latency with high precision and recall. Skilled in Python, SQL, Power BI, and cloud technologies, with a solid foundation in supervised learning and neural networks. Aiming to leverage expertise in applied ML and computer vision for impactful projects in safety-critical and enterprise environments.

• Machine Learning Specialization — Stanford University and DeepLearning.Ai
• IBM Data Science Professional Certificate — IBM
• Google Cloud Digital Leader — Google
• Bi dashboards with Tableau — Codeacademy
• Bi dashboards with Power BI — CodeAcademy
• TCS iON Career Edge – Young Professional – TCS iON

• PRISONSECURE: Smart Prison Surveillance (YOLOv11, FaceNet+SVM, OpenCV, Python)
  Designed and implemented a real-time surveillance system that detects weapons from CCTV feeds and recognizes inmate identities to trigger immediate alerts and evidence logging. Fine-tuned YOLOv11 with advanced augmentation (mosaic, mixup, RandAugment, erasing), focal loss, and label smoothing to improve precision/recall; built a FaceNet embedding pipeline with an RBF-SVM classifier for robust recognition under varied lighting and partial occlusion. Engineered a threaded audio alerting mechanism, implemented confidence thresholds, and documented design trade-offs, metrics, and error analysis to support near real-time operation and 24×7 stability.
• ASL Live Image Recognition (CNN/Transfer Learning, OpenCV, Python)
  Developed a webcam-based application that classifies American Sign Language gestures in real time and translates them to English text on-screen. Built a preprocessing and segmentation pipeline for frames, trained a CNN/transfer learning classifier on labeled signs, and added confidence smoothing to stabilize predictions. Delivered a functional prototype within a month post-training, demonstrating dataset preparation, rapid iteration, and user-visible inference loops for accessibility use cases.
• IBM Data Science Capstone: SpaceX Falcon 9 Landing Prediction (Python, SQL/DB2, Plotly Dash, Folium, scikit-learn)
  Built an end-to-end ML workflow to predict first-stage landing success and inform launch cost competitiveness. Collected mission data via SpaceX REST API and Wikipedia web scraping; performed wrangling with one-hot encoding and missing-value imputation; engineered a binary target for landing outcomes. Conducted EDA in Python and SQL, built Folium maps to analyze site proximities, and created a Plotly Dash app to explore success by launch site, orbit, payload, and year. Trained and tuned Logistic Regression, SVM, KNN, and Decision Tree models with GridSearchCV; evaluated on a hold-out set with accuracy and confusion matrix, and summarized actionable insights and recommendations.
• ATS-Friendly Resume Optimizer (Vertex AI, RAG/Embeddings, Python, NLP)
  Built an application that compares a candidate’s resume with a target job description to score coverage, diagnose gaps, and generate prioritized recommendations for content and formatting. Implemented an embedding-based similarity layer with retrieval-augmented prompting on Vertex AI, standardized outputs via prompt templates, and set up evaluation loops to monitor consistency and quality. Delivered an ATS-focused analyzer that surfaces missing keywords, alignment suggestions, and structure fixes to improve recruiter screening and parsing.