Gneya P

Major Projects

Multi-Agent Recruitment Document Generation System – Automating End-to-End Hiring Documentation

Built a multi-agent AI system to automate recruitment document generation from team discussions and meeting notes. Developed a sequential workflow using CrewAI to manage agents responsible for scorecards, job descriptions, and outreach messages. Integrated Retrieval-Augmented Generation (RAG) with OpenSearch for semantic document retrieval. Engineered human-AI collaboration via markdown-based reviews and implemented context-passing, version control, and PDF output pipelines using Flask and PostgreSQL for high-quality, consistent recruitment documentation.

USA

Frontend:

HTML5 & CSS3
JavaScript

Backend:

Flask
Python

Database:

PostgreSQL

AI/ML:

CrewAI
Embedding models
LLM integration

Automated Document Intelligence – Smart Classification and Structuring

Designed an end-to-end AI system to classify, extract, and structure data from various document types. Utilized NLP and GenAI for accurate field extraction and document recognition. Automated workflows using LangChain and OpenAI, with AWS integration for secure cloud operations. Applied JSON transformation for downstream compatibility and implemented robust validation for data accuracy.

UK

Backend:

FastAPI
Python

AI/ML/NLP:

LangChain
OpenAI GPT
Paddle

Disability Benefits Doc Automation Platform – AI-Powered Assistance for SSI/SSDI Applications

Developed an end-to-end platform to simplify SSI/SSDI applications with AI-driven tools like a Conversation AI chatbot for form assistance, Document OCR for data extraction, and Audio Transcription for voice-to-text conversion. Built prompt-tuned LLMs for accurate support, implemented a Flags and Remarks system to ensure data accuracy, and integrated MongoDB and PostgreSQL for scalable and reliable data handling. Focused on usability, precision, and compliance through continuous testing and refinement.

USA

Database APIs:

FastAPI
MongoDB
PostgreSQL

AI/ML:

LangChain
OCR
Whisper for STT

Automated Document Verification System – OCR-Powered Form Validation

An intelligent system for validating user-filled forms against structured datasets, leveraging OCR to extract text from scanned images and comparing it with tabular data (e.g., Excel) using Word Error Rate (WER) and Character Error Rate (CER) metrics. Configurable thresholds are determined from acceptance or rejection, enabling automatic flagging and reducing the need for manual verification. The system streamlined document processing, improved accuracy, and enhanced operational efficiency in high-volume environments.

India