AI Systems · Backend · Infrastructure · Performance
Computer Science, KJ Somaiya School of Engineering · Mumbai, India · Class of 2028
Overview · Awards · Experience · Projects · Open source · Toolbox · Learning frontier · Contact
I’m Shantanav, a computer science student and engineer interested in what happens behind useful AI products: agent workflows, retrieval and evaluation, backend services, data movement, and the runtime that makes software fast and reliable.
I’ve built multi-service AI products, contributed fixes to established open source projects, and worked on a Rust and WebAssembly rendering pipeline that reduced render latency by 40%. I’m deliberately going deeper into core machine learning, operating-system internals, networking, and performance engineering.
| 40% lower render latency |
7 merged upstream PRs |
4 open contributions |
3 hackathon podium results |
| Area | The work I’m drawn to |
|---|---|
| AI systems | Agent orchestration, retrieval pipelines, asynchronous inference, and evaluation |
| Backend and infrastructure | Services, queues, databases, event flows, deployment, and observability |
| Systems and performance | Rust, WebAssembly, GPU execution, Linux, and understanding costs below the API |
| Product engineering | Turning an idea into a usable tool with a clear interface and dependable behavior |
| Result | Event | Project or track |
|---|---|---|
| 1st place | ETH Mumbai 2026 | BitGo DeFi · VCR Protocol |
| 2nd place | ETH Mumbai 2026 | Base AIx Onchain · VCR Protocol |
| 2nd place | Solana Blitz V7 | Among 300+ international teams |
The two ETH Mumbai results were earned with Team OmShantyOm for VCR Protocol.
August 2026–present
I build AI workflows that connect conversational interfaces to business operations.
- Implemented WhatsApp agent workflows with SagePilot, including intent routing, structured tool calls, and backend API integrations.
- Integrated Razorpay payment initiation, transaction verification, and status handling into the workflows.
- Worked across the product boundary: translating operational needs into agent behavior and dependable service interactions.
June–August 2026
I worked on the video pipeline behind browser previews and exported videos.
- Built a shared Rust and WebAssembly rendering core for browser previews and backend exports, reducing preview and export drift.
- Added GPU parallelization to the rendering path, reducing render latency by 40%.
- Worked across Rust, WebAssembly, GPU execution, and the application surface that uses the renderer.
How the shared rendering path fits together
Both preview and export use the same rendering core. The core can use a GPU path when available and a CPU fallback when it is not.
flowchart LR
preview["Browser preview"] --> core["Rust and WebAssembly core"]
export["Backend export"] --> core
core --> gpu["GPU path"]
core --> cpu["CPU fallback"]
I enjoy working in established codebases where a small, well-tested change can improve behavior for many downstream users. The snapshot below reflects the PRs listed here; statuses were checked on 26 September 2026.
Merged upstream · 7 pull requests
- #1257 Retain usage metadata when later streaming chunks omit it.
- #1258 Propagate external cancellation through workflow execution.
- #1259 Respect the configured isolation scope for remote agents.
- gVisor #14083 Include a generated changelog in the Debian
runscpackage. - ORAS #2184 Publish stable rolling tags for container releases.
- Hugo #15148 Resolve the
staticcheckexecutable aftergo install. - Vello #1807 Document public Cargo features.
Open contributions · 4 pull requests
- ORAS #2191 Add multi-platform selection to
oras cp, including referrer preservation for the selected platform. - kind #4268 Add a local Markdown lint target and changed-file CI workflow.
- Kubeflow Testing #1087 Add a reusable Kind-cluster GitHub Action for test workflows.
- Google ADK Go #1256 Recover stale database sessions, with a two-writer regression test.
Browse all my public pull requests
| Area | Tools and concepts |
|---|---|
| Languages | Python, Go, Rust, TypeScript, JavaScript, C, C++, Dart, Solidity |
| AI and retrieval | LangGraph, RAG, embeddings, vector search, agent workflows, evaluation |
| Backend | FastAPI, NestJS, Node.js, asynchronous workers, REST, gRPC |
| Data | PostgreSQL, Redis, MongoDB, Qdrant, Chroma |
| Systems and infrastructure | Linux, Rust and WebAssembly, WGSL and WebGPU, Docker, Kubernetes, GitHub Actions |
| Product surfaces | React, Flutter, TypeScript |
I’m happy to talk about AI systems, backend and infrastructure work, or a concrete open source issue.