Victor I.

Owned end-to-end delivery of the MVP of a social platform connecting creators with brands, enabling discovery, collaboration, and campaign coordination between both parties, including creator-brand matching and engagement workflows from onboarding to active collaboration.

On the engineering side, I developed backend APIs, system integrations, and handled AWS cloud deployment for the mobile application. I also enabled fast iteration through CI/CD deployments and automated rollbacks improving stakeholder testing and feedback cycles during early product validation.

Alongside core development, I supported the team by mentoring junior developers through code reviews and technical guidance, which helped improve code quality consistency and reduced PR iteration time by around 20%.

• • Developed conversational AI services using LLM-based pipelines integrated with real-time com munication frameworks to support low-latency voice interactions and automated workflows.

• • Stabilized speech and agent execution workflows across text-to-speech (TTS) and speech-to-text (STT) systems by implementing asynchronous service patterns and message queue architectures to improve reliability and responsiveness.

• • Supported production deployments by working alongside senior DevOps engineers to implement monitoring, health checks, and rollback procedures, improving system stability and reducing service disruptions under live traffic conditions.

Worked as part of a team to develop real-time computer vision and NLP pipelines for humanoid robotic systems, including object detection, tracking, pose estimation, and speech-based interaction components. Contributed to dataset preparation and augmentation at scale, including structured labeling for vision and language datasets to support model training and evaluation. Supported implementation and integration of YOLO/CNN-based inference pipelines for real-time perception, alongside NLP components for speech-to-text processing, intent classification, command parsing, and dialogue logic. Evaluated system performance using metrics such as accuracy, F1 score, mean average precision (mAP), latency (ms), and frames per second (FPS), while analysing failure cases including false positives, edge-case behaviours, and model drift to improve robustness in deployment. Assisted in deploying models onto robotic hardware and conducting real-time testing in constrained edge environments, optimising for latency, memory usage, and responsiveness.