About me

Why I’m Different ● Generative AI First – Fine-tune LLMs, Stable & Video Diffusion, and integrate them into products that users love. ● Real-Time AI Agents – Architect low-latency voice/chat avatars on WebRTC & WebSocket, handling thousands of concurrent sessions. ● Scale Without Pain – Orchestrate distributed PyTorch across multi-node GPU clusters for 40% faster convergence and lower cost. ● DevOps DNA – Everything ships container-first (Docker) and self-heals in Kubernetes with automated CI/CD pipelines. Proof in Practice: ● Upgraded a voice-AI platform to v1.x, adding RAG, memory, and n8n workflows; rolled out via deployments on K8s. ● Delivered sub-200 ms talking avatars, pruning Wav2Lip and streaming through LiveKit, slashing GPU spend by 50%. ● Trained Stable Video Diffusion on 4×A100 & 4×H100 GPUs, integrating ControlNet, IP-Adapter & 3D U-Net for faster convergence. ● Built a real-time election analytics pipeline (Debezium → Kafka → PySpark) that streamed live results for 63 regions. ● Published ExoSGAN/ExoACGAN, achieving perfect recall & precision in exoplanet detection. Core Stack Python · C++ · Java | PyTorch, LangChain | Kafka, Spark | AWS & GCP | Docker, Kubernetes | WebRTC, REST, gRPC | CI/CD (GitHub Actions)