Piyush Choudhari

Pune, India AI & Backend Engineer LinkedIn GitHub Resume choudhari.piyush@gmail.com

About Me

I’m an engineer who enjoys building systems that don’t just work—but scale, explain, and solve real-world problems. From developing AI-powered products to architecting backend infrastructures, I focus on creating solutions that balance complexity with clarity. I’m passionate about working in teams that value thoughtful problem-solving. If the work challenges me to think deeper and build smarter, I’m all in.

Skills

Experience

Ronin Labs – AI Engineer Intern (Jan 2025 – Jul 2025)
Multimodal Content Generation System
- Built an agent-based orchestration backend using LangGraph + OpenAI for routing image/video/audio tasks.
- Implemented SDXL-based image outpainting and MusicGen-based audio generation.
- Optimized GPU task throughput under 20s latency.

Sketchbot Link
- Engineered FastAPI + SDXL backend to generate and optimize line-art G-code for Dexarm.
- Achieved 70% reduction in sketch time with optimized pipelines.
- Deployed on AWS g5.2xlarge, secured Raspberry Pi auth, and automated disaster recovery.

OnePlus 13s Quest Quiz Link
- Built scalable Node.js/TypeScript backend for 150K+ users, with <120ms API latency.
- Developed Django admin panel and managed Cloud SQL on GCP for global-scale ops.

Projects

KnowFlow
GitHub | Demo | HLD
- Hybrid document retrieval using PGVector + Neo4j for multi-hop reasoning via Gemini API.
- Scored ~75% Recall@5 with retrieval pipeline optimization and chunk feedback loops.
- Deployed on AWS ECS Fargate with CI/CD (GitHub Actions) and CloudWatch logging.
TrackML
GitHub | Demo
- Full-stack model tracker integrating Gemini + Groq LLaMA-4 for summarizing research URLs.
- FAISS vector search over Neon-hosted PostgreSQL, achieving 200ms semantic query latency.
- Deployed on EC2 with Flask + Nginx + GZIP, serving 100+ QPS with custom API instrumentation.

Education