Aswin PB

AI Researcher

Kochi, Kerala, India0 mo experience

AI EnabledAI ML Practitioner

Key Highlights

Developed advanced AI systems for real-time object detection.
Strong foundation in AI technologies and full-stack applications.
Proficient in multiple programming languages and frameworks.

Stackforce AI infers this person is a skilled AI/ML developer with expertise in computer vision and full-stack application development.

Contact

Skills

Core Skills

Machine LearningComputer Vision

Other Skills

Remote DevelopmentPythonPyTorchOpenCVsupervisionModel Context Protocol (MCP)Vector DatabasesLlamaIndexRetrieval-Augmented Generation (RAG)MongoDB atlasNLUspaCyrasaResearch SkillsData Structures

About

I am a passionate AI and Machine Learning enthusiast with hands-on experience in developing intelligent systems and full-stack applications. My academic and project journey reflects a strong foundation in both theoretical concepts and practical implementation of AI technologies. • AI/ML Projects: • Developed a deep learning model for the classification of Indian art forms, focusing on image-based recognition using neural networks. • Executed a real-time object detection system using a Raspberry Pi, integrating computer vision techniques and hardware-level optimization for efficient edge computing. • Software Development: • Proficient in Python, C, C++, Java, Shell Scripting, and PHP. • Developed a car rental website using PHP, showcasing server-side scripting and database interaction skills. • Prompt Engineering: • Applied prompt engineering techniques to design effective interactions with large language models, enhancing AI output quality and task performance. • Strengths: • Strong debugging and problem-solving skills across different programming environments. • Effective at bridging AI models with practical deployment using both software and embedded systems. I enjoy creating solutions that combine machine learning intelligence with robust and scalable application architectures.

Experience

0 mo

Total Experience

Average Tenure

Current Experience

Uptoskills

Deep Learning Intern

Jul 2025 – Oct 2025 · 3 mos · Remote

Role: Deep Learning Intern, Uptoskills (CSS Crowd Surveillance Team)
As part of the Crowd Surveillance Systems team, I developed an end-to-end AI pipeline for the real-time detection of chain-snatching events from video. The system efficiently identifies high-risk interactions by intelligently filtering video data, avoiding the high computational cost of constant analysis.
My Contributions:
Data & Detection: I contributed to the collection and preparation of a custom video dataset for training and validation. I then trained and fine-tuned a YOLOv8 model for high-accuracy detection of key entities (person, bike_scooter).
Tracking System Upgrade: I engineered a robust multi-object tracking system, initially using DeepSORT. I later upgraded this critical module to the state-of-the-art ByteTrack algorithm, which significantly improved tracking performance through occlusions and increased system speed by eliminating the need for a separate Re-ID model.
Action Recognition & Heuristics: I integrated and benchmarked multiple 3D-CNN architectures (C3D and I3D) for the final action classification. I developed a proximity-based trigger to intelligently activate the I3D model only during suspicious person-person interactions. Furthermore, I implemented a post-event velocity heuristic to identify and persistently track the likely perpetrator, enhancing situational awareness.
Technical Workflow:
The pipeline operates in a cascade: YOLOv8 detects, ByteTrack tracks, a proximity trigger activates analysis, and I3D classifies the event.
Technologies Used:
Languages & Frameworks: Python, PyTorch
Computer Vision Libraries: OpenCV, supervision
Core Models: YOLOv8, ByteTrack, I3D, C3D