Aaditya Barak

Aaditya Barak

AI Engineer — LLMs, RAG Systems, Document Intelligence

• Phone: +91 9896127290
• Email: aadityabarak970@gmail.com
• LinkedIn: linkedin.com/in/aaditya-barak
• GitHub: github.com/AddyB0t

Technical Skills

• Languages: Python, SQL, JavaScript
• AI/ML: PyTorch, Transformers, NLP, Computer Vision, OCR
• LLM & GenAI: LLMs, Prompt Engineering, RAG, LangChain
• Databases: PostgreSQL, Supabase, Redis
• Tools: Git, Linux

Experience

PrudentBit — Artificial Intelligence Engineer

May 2025 – December 2025 | Noida, Uttar Pradesh

• Led development of a production-grade document redaction system for PDF, DOCX, images, reducing manual effort by 70% across 1,000+ documents
• Built an end-to-end document AI pipeline (OCR + preprocessing + PII detection + masking), improving processing speed by 2x
• Designed preprocessing pipeline improving OCR accuracy by 25–35%
• Developed PII detection system with pixel-level masking while preserving layout
• Implemented TinyBERT-based reranking (100ms) improving search relevance

Freelance AI Engineer — Self-Employed (Client Projects)

May 2025 – Present | Remote

• ZapUp – AI Educational Platform: Built full-stack platform with real-time chatbot, personalization, and scalable backend using Next.js and Supabase
• Implemented context-aware LLM assistant with conversation memory and dynamic filtering
• TITTA – AI Construction Analysis System: Built production-grade RAG system for querying PDFs and CAD (DWG/DXF)
• Implemented vector search using BGE-large embeddings with Supabase vector storage
• Designed DWG/DXF metadata extraction pipeline with structured storage in PostgreSQL
• Enabled domain-specific reasoning and automated workflows for engineering analysis

Projects

AI-Powered Resume Parser & Job Matcher

FastAPI, PostgreSQL — November 2025

• Built AI system using OCR + NER + LLM scoring processing 1,000+ resumes
• Implemented multi-criteria ranking reducing recruiter effort by 40%
• Designed scalable FastAPI backend with parsing and scoring pipeline

Deepfake Detection System

TensorFlow — April 2025

• Achieved 93% accuracy using VGG16 transfer learning
• Evaluated performance across real vs synthetic datasets

BERT NER System

Transformers, PyTorch — July 2025

• Built token classification model using HuggingFace Transformers and PyTorch
• Improved performance using fine-tuning and preprocessing

Education

UIET, Panjab University — Bachelor of Engineering (Electrical & Electronics Engineering) Sep 2022 – May 2026 | Chandigarh, India

Leadership / Extracurricular

Open Source and Developer Community — Contributor / Mentor (2022 – Present)

• Contributed to ML/NLP repositories and mentored developers on APIs and deployment