Kategorien durchsuchen
Entdecken
Fiverr Pro
Deutsch
$
USD
Mani Varun
@manivarunarivin
Gen AI Developer
Indien
Englisch, Deutsch, Telugu, Hindi
Über mich
I am a Gen AI, computer vision and deep learning engineer passionate about automating workflows and deploying AI at scale. Currently, I am working at Piresearch Labs Pvt. Ltd as Generative AI developer building AI solutions which include RAG system, Chatbots, and OCR. Recently, I had completed my Master's thesis at Leibniz IPHT in Jena, Germany, where I had developed a cloud-based data fusion and object detection pipeline using Nextcloud, Python, ONNX, and Shiny for real-time remote image analysis.
Kompetenzen
Mani Varun
offline •
Meine Dienstleistungen
KI-Technologie-Beratung
I will develop rag genai and computer vision solutions
Arbeitserfahrung
Gen AI Developer
Piresearch Labs Pvt. Ltd. • Vollzeit
Jul 2025 - Present • 10 mos
• Project 1: Built and deployed a Retrieval-Augmented Generation (RAG) system on an AWS EC2 instance, integrating Google Gemini API, FAISS, and AWS RDS (PostgreSQL) for scalable semantic search and FAQ retrieval. • Engineered data pipelines in Python, securely managing credentials via AWS Secrets Manager, storing metadata in PostgreSQL, maintaining FAISS indices in Amazon S3, and automating index refresh using AWS Lambda triggered EC2 workflows. • 2. Support Module: Developed a client-facing ticketing system with P1–P4 priority classification, SLA tracking, and dedicated Slack alerts, including a separate escalation channel for P1 (critical) tickets. • Current Work: Developing a custom & scheduled reporting system by defining standardized reportable data schemas. Implementing dynamic query execution to support both scheduled and on-demand reports.
Leibniz IPHT, Jena, Germany
Vollzeit • 1 yr
Master Thesis in Remote Object Detection System with Raspberry Pi, YOLOv8, and Python Shiny
Jan 2025 - Jul 2025 • 6 mos
• Developed an end-to-end cloud pipeline integrating Nextcloud, Python-based processing, and Shiny App for real-time data fusion and predictive analytics. • Engineered a refresh-button triggered inference system that performs YOLOv8s ONNX predictions exclusively on newly captured images, using a tracking file. • Enabled remote configuration of Raspberry Pi camera through a shared config.json. • Implemented dynamic camera modes (Idle, Single Shot, Timed Loop) to provide flexible and automated image capture control from Shiny dashboard. • Reducing model initialization time from 23s to 5s and inference time per image from 3s to 0.5s using ONNX Runtime.
Mandatory Internship in Deep Learning for Image Classification and Object Detection using Transfer Learning at Leibniz IPHT, Jena, Germany
Jul 2024 - Jan 2025 • 6 mos
• Designed and built a Raspberry Pi-based Microscope to capture and annotate a custom leaf image dataset (219 images across 10 classes) for computer vision applications. • Enhanced model performance through data augmentation techniques (including rotations and flips) and hyperparameter optimization. • Achieved 88% classification accuracy using AlexNet and 94.9% mAP@0.5 with YOLOv8 on the test set by leveraging transfer learning and advanced hyperparameter tuning. • Successfully deployed optimized models to the Raspberry Pi Microscope using ONNX for real-time inference and predictions. • Tools & Technologies: PyTorch, Linux, Python, ONNX, OpenCV, Picamera2, LabelImg.
