Surya P.

Key Skills

Software

No software listed

Top Subject Matter

US address registry and document Q&A

Image-based quality classification and document/vision processing automation

Top Data Types

Document

Image

Top Task Types

Question Answering

Data Collection

Freelancer Overview

AI/ML Intern — DataFactZ Solutions Pvt. Ltd. (Document RAG Q&A + address extraction). Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include ChromaDB, Google Cloud Vision, and NLP. Education includes Bachelor of Technology, Aditya College of Engineering and Technology, JNTUK (2026). AI-training focus includes data types such as Document and labeling workflows including Question Answering and Data Collection.

Labeling Experience

AI/ML Intern — DataFactZ Solutions Pvt. Ltd. (Document RAG Q&A + address extraction)

DocumentQuestion Answering

Built and evaluated a document-based RAG Q&A system that performs deterministic document address extraction plus LLM-assisted normalization and retrieval. Tracked retrieval and refusal-quality metrics including Recall@4, MRR, and refusal rate using an automated evaluation scorecard. Integrated outputs into production-grade FastAPI services for API consumption. • Created a document pipeline (upload → extract → normalize → dedup → embed → retrieve → rerank → ask) • Implemented CrossEncoder reranking with a refusal-threshold guardrail and query rewriting • Designed evaluation runs for Recall@4, MRR, and refusal rate reporting • Delivered results through FastAPI endpoints and supporting test automation (pytest)

2026 - Present

AI & ML Engineer Intern (APSCHE Govt. Certified) — SmartBridge / APSCHE (AI-powered apps + classifier)

DocumentData Collection

Developed AI-powered web applications and ML pipelines that streamline processing tasks by automating supervised and deep learning inference. Integrated external Google Cloud Vision/NLP/AutoML services to support model-driven workflows. Delivered a production-ready CNN classifier and demonstrated end-to-end deployment for real-time use. • Built 3 AI-powered web applications (Python, TensorFlow, Streamlit) reducing manual processing time by ~40% • Implemented supervised and deep learning pipelines achieving >88% accuracy • Integrated Google Cloud Vision/NLP/AutoML via reusable SDK wrappers • Deployed an AI-powered fruits quality CNN classifier with 94%+ accuracy in Streamlit

2025 - 2025

Education

Work History