Stephen M.

Key Skills

Software

Other

Top Subject Matter

Handwritten text transcription / OCR

GPS spoofing detection / signal-based classification

Biometric retrieval / forensic image-to-identity ranking

Top Data Types

Text

Top Task Types

Prompt + Response Writing (SFT)

Freelancer Overview

Handwritten Text to Word Document Conversion project (TrOCR-based OCR).. Core strengths include Other. Education includes Bachelor of Technology, Ace Engineering College (2026). AI-training focus includes data types such as Computer Code and Programming and labeling workflows including Prompt + Response Writing (SFT), Evaluation, and Rating.

Labeling Experience

Forensic Sketch Generator project (composite generation + biometric matching).

Other

The resume describes an AI forensic pipeline that generates suspect composites and performs biometric matching from a mugshot database. This includes using computer vision components for face embeddings and segmentation-style parsing, then ranking suspect matches by composite similarity to support downstream decisions. No explicit labeling job is stated; however, the work is consistent with building supervised or benchmark-based model workflows for biometric retrieval and ranking. • Built a forensic AI pipeline integrating Stable Diffusion XL for suspect composite generation. • Implemented a hybrid FAISS + InsightFace + BiSeNet matching system for ranked biometric retrieval. • Added multi-modal biometric scoring using facial embeddings, HSV skin tone analysis, and semantic hair/beard parsing. • Deployed a FastAPI GPU backend with an interactive dashboard and real-time forensic reports.

2026 - Present

GPS Spoofing Detection & Prevention System project (ML training + evaluation).

Other

The resume describes development of a GPS spoofing detection model along with a real-time evaluation dashboard. This entails working with GNSS signal feature inputs, training classical machine learning classifiers, and computing evaluation metrics such as Precision, Recall, and F1-score. While this is not described as data labeling, it reflects supervised model training and evaluation over labeled attack/normal scenarios. • Trained Random Forest, Decision Tree, and Naive Bayes classifiers for GPS spoofing detection. • Built a Streamlit dashboard to display Precision, Recall, and F1-score during evaluation. • Analyzed signal anomalies to improve detection performance across attack scenarios. • Used Scikit-learn for model training and metric computation.

2026 - Present

Handwritten Text to Word Document Conversion project (TrOCR-based OCR).

OtherPrompt + Response Writing (SFT)

The resume describes AI development work focused on OCR and handwriting-to-text transcription using a Transformer-based TrOCR pipeline. This involves preparing textual inputs/outputs, designing preprocessing and post-processing stages, and validating transcription formatting for editable document outputs. No explicit human labeling or dataset annotation role is stated, so the experience is inferred as AI training/prompt-to-output pipeline implementation for transcription. • Implemented Transformer-based OCR to convert handwritten text into editable digital documents. • Designed preprocessing/post-processing to improve transcription accuracy. • Ensured formatting consistency for the resulting digital documents. • Used TensorFlow and Keras for the OCR model workflow.

2026 - Present

Education

Work History