Open to opportunities

Pushpak Hinglaspure

@pushpakhinglaspure

Message

I build GPU-optimized, production generative AI pipelines with multi-agent capabilities.

India

Message

What I'm looking for

I’m looking to build production generative AI with real GPU throughput gains—GPU inference optimization, multi-agent pipelines, and strong MLOps/observability in Azure/AWS teams that value reliability and speed.

I’m an AI/ML Engineer specializing in GPU inference optimization and production-scale generative AI. I achieved 85% inference speedup on H100 clusters and architected a zero LangChain multi-agent video pipeline driven by a locally-served 72B-parameter VLM.

In my current role, I cut compute by adapting denoising step inference (about ~50%) and optimized inference serving with Triton, TensorRT, FlashAttention-3, and PyTorch Compile, scaling GPU concurrency from 1 to 4 simultaneous generations. I’ve also led end-to-end cloud migration (AWS → Azure) and built reliability-focused tooling like TSX self-repair, plus a production observability dashboard tracking GPU/VRAM/CPU load and PCIe bandwidth across H100 clusters.

Experience

Work history, roles, and key accomplishments

Current

Senior AI/ML Engineer

Current

Phenomenal AI Pvt. Ltd.

Aug 2025 - Present (9 months)

Optimized GPU inference for a 22B proprietary diffusion model, improving latency from 2 min to 45s on a single H100 (62.5% faster) while maintaining zero quality loss. Architected a zero-LangChain agentic video pipeline and delivered cross-cloud migrations from AWS to Azure, scaling concurrent GPU generations from 1 to 4.

PyTorch Triton Inference Server TensorRT VLLM Kernel Fusion Azure AWS EC2 Remotion

Applied AI Founder

Persist Ventures

Aug 2024 - Sep 2025 (1 year 1 month)

Founded VidGenCraft, driving 1,000+ users, 75% engagement, and 95% cost reduction, and led social automation for 50+ accounts with 24/7 operations. Built high-throughput content automation processing 21,000+ posts/month (500M+ views) using an async distributed AWS architecture and delivered AI analytics and monitoring dashboards.