Work Experience

Professional Journey & Technical Achievements

June 2023 - August 2025

Opkey

Noida, India

AI Software Developer (Part-time + Full-time)

Comprehensive AI development role spanning from internship to full-time position, focusing on enterprise automation, LLM optimization, and scalable AI solutions for ERP systems.

Key Technical Achievements

Designed and implemented a classifier-based architecture within multi-agent LLM workflows for ERP automation, achieving a 40x reduction in token-related costs and 75% overall decrease in processing expenses.
Integrated streaming capabilities into LLM pipelines, achieving an 85% reduction in latency. This major update was pivotal for client demos and directly enhanced solution responsiveness.
Architected and implemented a RAG-enabled backend AI architecture, reducing model size from 235B to 8B parameters and cutting operational costs by 95.83% while maintaining accuracy and enhancing scalability.
Built an AI-powered automation pipeline to generate 4,000+ product descriptions, reducing several weeks of manual work to just a few days at a cost of only $1.25.
Engineered a multi-class few-shot orchestrator achieving 94% accuracy, capable of asking clarifying questions and dynamically assigning agents based on identified use cases.
Fine-tuned BERT model for PII detection and classification with 95% accuracy, ensuring enterprise-grade data security compliance.
Engineered fake streaming mechanisms within LLM workflows to simulate responses during development, leading to an approximately 82% reduction in latency and enhanced developer efficiency.
Developed Generative AI-powered chatbot models capable of automatically generating test cases from natural language prompts, resulting in significant time and cost savings across QA workflows.
Implemented Chain-of-Thought reasoning for agent-based tool selection using sequential and dynamic I/O referencing, enabling accurate and context-aware responses to user queries.
Developed an LLM-based summarizer that converts raw text into well-structured HTML pages, eliminating the need for customers to parse JSON or lengthy Excel files.
Technologies & Tools
Python FastAPI Flask LangChain LlamaIndex Hugging Face OpenAI API Qdrant Redis Docker LM-Studio Ollama Langfuse BERT
Skills Developed
LLM Hosting Multi-Agent Systems RAG Architecture Cost Optimization Latency Reduction Team Leadership Training & Mentoring ERP Automation