Junior AI Engineer at Vertex Special Technology

Focused on Large Language Models and NLP domain. Implemented OpenAI key for OCR-based project to map extracted OCR data into required JSON format. Studied transformer architectures (Mistral-7B, Llama, Gemma models) and implemented, tested these models on cloud-based GPUs. Gained knowledge of fine-tuning methods like PEFT, LORA, QLORA, DPO.

Trainee AI Engineer at Vertex Special Technology

Built Tele-plant Doctor mobile application for FAO project detecting plant diseases from leaves. Implemented YOLO v7 and v8 models for object detection. Gained expertise in object detection, data annotation (bounding boxes), and data cleaning. Worked effectively with team on assigned tasks.

AI Engineer - Intern at Vertex Special Technology

Worked on various machine learning and deep learning algorithms. Conducted in-depth research on OCR algorithms including Tesseract, Paddle OCR, and Easy OCR. Developed strong Python programming skills and understanding of OOP concepts.

AI Engineer at Braintip AI

Architected a multi-agent workflow automation system translating natural language queries into executable, database-driven workflows with support for complex conditional logic, loops, and 20+ event types. Built a dynamic form routing engine that extracts trigger/target forms and maps event codes to action agents without hard-coded dependencies, reducing workflow creation time by 90%. Designed modular agent pipeline (planner→orchestrator→classifier→mapper→optimizer) to handle complex scenarios. Implemented robust JSON generation system for CRUD operations, email/SMS notifications, user management, and data transformations.

Associate AI Engineer at Vertex Special Technology

Contributed to Version 1 release of an OCR-based accounting automation system using PaddleOCR to extract structured and unstructured financial data and map it into validated JSON for downstream accounting workflows. Gained architectural knowledge of transformer architectures including Mistral-7B, Llama, and Gemma; tested on cloud-based GPU for production use cases. Practiced fine-tuning methods including PEFT, LORA, QLORA, and DPO; gained expertise in agent-based systems using LangGraph, LangChain, and CrewAI frameworks. Researched Vision-Language Models (VLMs) with a focus on Qwen VLM to improve document understanding. Fine-tuned and trained Qwen VLM models for Proof-of-Concept, achieving improved extraction accuracy on complex financial documents and validating feasibility for production. Integrated ASR for voice-driven actions in accounting workflows.