π Currently building Vigoe AI β a real-time multi-tenant voice AI calling platform with WebSocket audio pipelines, adaptive NLMS echo cancellation, and RAG integration. π― Collaborating on Computer Vision (Materials Science) and GenAI pipelines. π± Exploring On-device AI segmentation and seamless texture generation. π¬ Skilled in AI Voice Agents, Image Processing (OpenCV), Stable Diffusion, and FastAPI backends. β‘ Former Machine Vision Engineer Intern at Insitumicron β microscopic grain boundary analysis.
- Amyotrophic Lateral Sclerosis Disease Detection Through Facial Expression Analysis β IEEE ICAIHC, Jan 2025
- Image Caption Generation Using Contrastive Language Image Pretraining β IEEE ICCCNT, Jun 2024
- Comparative Study of EfficientNet and MobileNet Models for Lung Cancer Classification Using Chest CT Scan Images β IEEE ICETITE, Feb 2024
- Fake News Detection Using a Stacked Ensemble of Machine Learning Models β IEEE IDCIoT, Jan 2024
- AI & Backend Developer (Home Ideas Technologies): Deployed real-time AI voice agents, n8n automation pipelines, and a computer vision AI Visualizer for interior design simulation β serving 15+ client companies.
- Machine Vision Engineer Intern (Insitumicron): Established grain boundary detection pipelines processing 500+ microscopy images monthly, cutting manual processing time by 40%.
- Data Analyst Intern (Odugaa Tech): Constructed end-to-end data pipelines and trained predictive ML models on 100K+ row datasets achieving 85% accuracy.
- Vigoe AI β Real-Time Voice AI Calling Platform: Multi-tenant AI calling platform with bidirectional WebSocket audio, adaptive NLMS echo cancellation, outbound campaign engine, and RAG-based knowledge retrieval. Stack: Python, FastAPI, WebSockets, asyncio, Redis, NumPy/SciPy.
- AI-Powered Query System: RAG pipeline extracting text from URLs, PDFs, and images (OCR) with Milvus vector storage and sub-second search. Stack: FastAPI, Milvus, Streamlit, Docker, Google Gemini AI.
- Fake News Detection β Stacked Ensemble ML: 94% accuracy on 72K+ samples using stacked Random Forest, SVC, Logistic Regression, and XGBoost. Published at IEEE IDCIoT 2024.
