Shubham K.

Key Skills

Software

Don't disclose

Top Subject Matter

Legal documents (legaltech) for RAG-based Q&A

Healthtech nutrition recommendations (RAG over user meal history)

E-commerce reconciliation (Amazon/Flipkart/Meesho) for RAG chatbot Q&A

Top Data Types

Text

Document

Top Task Types

Question Answering

Text Generation

Entity (NER) Classification

Freelancer Overview

Founder & Lead Developer — NyayX (LLM-powered document Q&A with RAG over embedded PDF chunks). Brings 17+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Don't disclose. Education includes Bachelor of Computer Applications, N/A (2009). AI-training focus includes data types such as Computer Code, Programming, and Text and labeling workflows including Question Answering, Text Generation, and Fine Tuning.

Labeling Experience

Founder & Lead Developer — NyayX (LLM-powered document Q&A with RAG over embedded PDF chunks)

Don't discloseQuestion Answering

Built an LLM-powered document Q&A system for a multitenant legaltech SaaS by chunking uploaded legal PDFs, embedding them, and storing vectors in pgvector. Implemented a RAG pipeline so users could query document content and receive answers grounded in retrieved chunks. Focused on enabling LLM question answering rather than traditional text labeling. • Data representation used: document chunks and vector embeddings in pgvector • Labeling/annotation work involved: preparing textual chunks for retrieval/Q&A • Software/components: OpenAI API, pgvector, S3 for documents, Redis caching • Application goal: tenant-isolated RAG-based Q&A over legal documents

2021 - Present

Founder & Lead Developer — DietGhar (AI nutrition recommendations with RAG)

Don't discloseText Generation

Integrated OpenAI API-based AI nutrition recommendations using a RAG approach over users’ meal-history vector stores. Engineered retrieval over stored meal-history embeddings so the system could generate recommendations grounded in user-specific nutrition context. Treated prepared meal-history textual data and embeddings as the core inputs for the retrieval-augmented generation flow. • Data representation used: meal-history vectors in pgvector • Labeling/annotation work involved: organizing/structuring user meal history for embedding and retrieval • Software/components: OpenAI API, pgvector, Next.js SSR/SSG frontend • Application goal: LLM-assisted nutrition recommendation feature

2019 - Present

Freelance Full Stack Developer and DevOps Consultant - Independent

TextEntity (NER) Classification

Worked as an independent full stack developer and DevOps consultant delivering production APIs, frontend applications, and cloud infrastructure for early-stage startups. Led implementations across e-commerce automation, AI/ML-enabled features, and scalable backend systems using modern queues and integration patterns. Operated across the product lifecycle including API design, authentication, CI/CD, and operational support for client deployments. • Built and maintained REST and GraphQL APIs with JWT/OAuth2 authentication and secure integrations • Developed scraper-based data ingestion systems and marketplace automation pipelines using Node.js and Python • Implemented queue/worker architectures with BullMQ and SQS and integrated messaging bots via WhatsApp/Telegram APIs • Applied AI/ML tools such as OpenAI API and RAG pipelines to deliver product features (non-training/data labeling).

2010 - Present

eVanik.ai Freelance (Ask eVa LLM chatbot with RAG over transaction data)

Don't discloseQuestion Answering

Developed an LLM powered e-commerce reconciliation chatbot (Ask eVa) using RAG for seller reconciliation discrepancy queries. Ingested seller transaction data into a pgvector vector store and built a retrieval pipeline to support natural-language questions about discrepancies. The core AI training/labeling-adjacent work centered on preparing structured transactional text for embedding and retrieval. • Data representation used: seller transaction records indexed into pgvector • Labeling/annotation work involved: transforming/ingesting transaction data into retrievable document chunks • Software/components: LangChain, OpenAI API, pgvector, Redis • Application goal: Q&A chatbot for reconciliation discrepancies across marketplaces

2024 - 2025

Education

Work History