FleekAI Software Engineer
Aug. 2025 - Dec. 2025United States- Developed advanced ComfyUI workflows enabling high-quality image generation, image editing, and robust LoRA training pipelines tailored for Flux Kontext model architectures, with seamless integration of state-of-the-art upscaling and detailing modules. - Conducted comprehensive research and rigorous benchmarking across leading Stable Diffusion models for image generation tasks—including Text-to-Image, Image-to-Image, and various editing use cases—delivering actionable insights that drove model selection and optimization. - Engineered a proof-of-concept integrating Nunchaku-based SVDQuant models, dramatically reducing image generation latency and GPU consumption compared to Flux and Qwen architectures. Achieved significant efficiency gains through comprehensive testing, profiling, and iterative enhancements. - Architected and deployed a scalable, multi-host observability platform utilizing VictoriaMetrics, VictoriaLogs, Grafana, vmalert, and Fluent Bit, all orchestrated within a declarative NixOS environment. Empowered real-time operational monitoring and granular log analysis across distributed AI infrastructure. - Enabled robust experimentation analytics by implementing Statsig within the FastAPI backend, delivering deep, actionable insights into feature rollouts, model performance, and A/B test results that informed rapid, data-driven decision-making.