Currently building Devin - cloud infrastructure for teams and AI agents to build software together.
DEVIN
A collaborative cloud workspace where teams work alongside AI agents to build, test, and ship software.
- 01Built for production teams with observability, evaluation, audit logs, cost controls, and organization-wide security.
- 02Teams assign software tasks to developers and AI agents, with shared context across the same codebase.
- 03Agents work in isolated cloud environments with controlled access to code, tools, secrets, and runtime state.
NextFlow
Apr 2026Visual AI Workflow Builder — a multimodal workflow engine combining text, image, and video pipelines into unified AI execution graphs.
- Built a multimodal workflow engine combining text, image, and video pipelines into unified AI execution graphs.
- Designed parallel execution + selective recomputation system, reducing latency and compute cost.
- Implemented workflow versioning, logs, and execution tracing for debugging and production reliability.
Orbyt
Feb 2026Unified API gateway for multiple AI providers with structured documentation, failover, and observability.
- Built unified API gateway for multiple AI providers (OpenAI, Gemini, Claude, etc.) with structured API documentation.
- Implemented failover + load balancing across models to ensure reliability.
- Designed tracing system for request debugging, performance monitoring, and observability.
Haven
Sep 2025College-exclusive dating platform with profile discovery and engagement-focused social features.
- Developed and launched a college-exclusive dating platform with profile discovery and engagement-focused social features.
- Sustained 100+ active users over multiple months.
- Achieved 17,000+ Instagram views in one month, driving organic traffic and user sign-ups.
- Worked with large open-source projects to create realistic software engineering challenges for AI coding agents (500+ GitHub stars).
- Designed coding challenges with automated tests, and reference solutions.
- Evaluated frontier AI coding agents on real-world debugging, feature implementation, and code reasoning tasks.
- Designed real-world software engineering tasks for evaluating frontier AI models.
- Created automated test suites and benchmarks to validate AI-generated code.
- Evaluated model outputs for correctness, reasoning, and software engineering quality.
- Led frontend development using Next.js, improving UI scalability and performance.
- Mentored peers via code reviews and debugging sessions.
- Delivered production-ready platform within 2-month deadline.
SRIJAN
I build, explore, and like to understand how things work.
I enjoy building things from scratch and understanding what happens underneath them. I started with products and software, which eventually led me deeper into AI, developer environments, autonomous agents, computer networks, and modern computer architecture.
Outside of code, I like reading novels, playing badminton, and exploring new technology.
- TypeScript
- Python
- Go · C++
- Node.js · FastAPI
- PostgreSQL · Redis
- Kafka
- Docker · Kubernetes
- AWS · Terraform
- GitHub Actions
- REST · gRPC
- WebSockets
- Linux
- LLMs · RAG · MCP
- Knowledge Graphs
- Agents · LangGraph
- LLMOps · PyTorch