Mano j Suresh

Engineering Intern at Newcastle University

Building agentic systems for different forecasting tasks, involving testing different frameworks and evaluating agent performance

Technical Consultant Intern at Indid Consulting Pvt. Ltd. (subsidiary of Dizparc Everest AB)

Collaborating with functional consultants and clients to analyze business requirements and translate them into technical solutions within the BC ERP ecosystem; designed and developed custom solutions using OOP, API integrations, and Azure cloud services; initiated and led independent AI research and built a Copilot capability to automate Excel data mapping and file imports; created an AI Agent to automate BC record insertion via MS Teams and Outlook

Research Intern at Alliance University Bengaluru

Developed DL models for banana plant disease detection using transfer learning with five different fine-tuned DCNN architectures; achieved highest test accuracy of 0.97 with the fine-tuned VGG16 model; awarded Best Paper at ICCCNet 2024

Civil AI Assistant at GitHub (Self-Initiated)

Built a multi-modal, voice-enabled assistant for the civil engineering domain using GPT-4o mini; applied Retrieval-Augmented Generation (RAG) pipeline on structural engineering manuals, design standards, case studies, and material property datasets; designed ETL-style pipeline to organize datasets for faster querying and analytics

Face Recognition Surveillance (Semi-Public Spaces) at Alliance University / GitHub

Developed a face recognition and entry management system for CCTV-based surveillance in semi-public spaces; compared multiple models (CNN-VGG, Viola-Jones, MobileNet, InceptionV2); designed a ResNet50 + ArcFace pipeline with accuracies 0.96 (1–5m), 0.80 (5–10m), and 0.60 in low-light conditions; centralized metadata storage with timestamps and camera locations

Adaptive Quiz using Unsupervised Learning at Alliance University / GitHub

Designed a Flask based quiz system that personalizes the quiz based on Bloom’s taxonomy levels and learner’s difficulty preference; managed data processing for training VAE model to predict question difficulty