
C2SI develops FOSS softwares for everyone!!!
Showing 5 of 26 projects. Click any project card for scope, mentors, and proposal studio.
Mentors: Student: Pranjal Udhwani
LLM agents built with LangGraph and LangChain retrieve documents, call tools, and write to memory - but nothing validates what enters the model's context. A poisoned RAG document can hijack an agent's behaviour. A malicious tool parameter can exfiltrate data. A single memory write can silently alter every future response. There's no standard security layer for any of this. ACF-SDK is a Zero Trust firewall that sits between the agent and the LLM - intercepting every input across four hooks (prompt, context, tool call, memory) and enforcing policy-as-code decisions through an isolated Go sidecar. I want to take this pipeline from hardcoded ALLOW to real enforcement: upgrade the semantic scanner to catch paraphrased attacks, wire scanner signals into the real sidecar, build memory write protection with HMAC and provenance tracking, ship a Docker demo anyone can clone and run in 5 minutes, and stress-test the whole system against adversarial evasion techniques with a technical evaluation report. The foundation is already built. I've shipped 6 PRs - semantic scanner, benchmark harness, LangGraph adapter, Rego policy templates for all four hooks, audit logger, and a Go scan stage implementation. 155 tests passing. GSoC wires these components into a running system and proves they hold up under real attacks.
Mentors: Student: Gautam Sidhwani
CodeLabz is an open source tutorial platform by C2SI built on React, Redux, and Firebase. The codebase has no TypeScript, uses deprecated MUI v4 patterns, React Router v5, and lacks key features like notifications and progress tracking. I will complete the TypeScript migration of the entire frontend and Cloud Functions, upgrade React Router v5 to v6, migrate MUI v4 styling to the modern sx prop, improve UI/UX across all pages, implement organization RBAC, build a real time notification system with Pub/Sub architecture, build an interactive quiz and progress tracking system, and optimize Firestore queries by adding missing composite indexes.
Mentors: Student: Chetan Barakki Satish Kumar
ImageLab is a block-based image processing tool built with Blockly and OpenCV . It helps students learn image processing by building pipelines with visual blocks. Problem statement: This project has the following key issues: This project will implement "Interactive Learning Mode" transforming ImageLab into a transparent, collaborative, and scalable learning ecosystem. First, per-step processing that returns intermediate image outputs after every operator, displayed in a scrollable filmstrip timeline so students can see the output after each step and what each step is contributing for the final output. Second, an image analysis panel showing RGB histograms and metadata for any selected step. Third, pipeline persistence allowing students to save, load, version, and share pipelines using secure share tokens. Fourth, a batch processing engine to run one pipeline across multiple images concurrently with progress tracking and ZIP download. Fifth, custom composite operators (macros) that let students save reusable block sub-chains into the Blockly toolbox. Deliverables: A production-ready fully tested interactive step viewer, image analysis panel with Canvas-based histogram renderer, SQLModel backed pipeline persistence with HMAC share tokens, asyncio batch processing engine, and a macro system with DAG validation and Blockly toolbox injection.
Mentors: Student: Atharva Tanaji Nagane
RustCloud unifies cloud provider APIs for Rust developers, but Generative AI support is completely missing. Every GenAI provider uses a different auth model, API style, and request format, forcing developers to maintain separate HTTP clients for each one. Rust has no equivalent of Python's LiteLLM or LangChain. This project delivers concrete LlmProvider implementations for AWS Bedrock, GCP Vertex AI, and Azure OpenAI, a complete GCP BigQuery module, a UnifiedLlmClient with Explicit/ModelBased/Fallback routing strategies, RetryMiddleware with exponential backoff, and a full test suite targeting 85%+ coverage with complete documentation. I have been contributing to RustCloud since March 2026 with 19 PRs across all four cloud providers, including working proof-of-concept implementations of all three GenAI providers and the routing layer. The GSoC period is where these drafts become production-quality, tested, and merged modules.
Mentors: Student: Udaybir Singh
GDB-UI currently supports only one user at a time because the backend shares a single global GDB controller across all requests. When two users debug simultaneously, the second session silently destroys the first with no error or warning. This proposal implements a three-phase solution. Phase 1 replaces the global controller with a thread-safe SessionManager that gives each user their own isolated GDB instance via a UUID session_id. A working implementation is already open as PR #131 with 15 passing tests, security fixes, and a live demo page. Phase 2 introduces WebSocket streaming with a per-session reader thread architecture so GDB output reaches the frontend in real time without blocking the server. Phase 3 moves each project into an isolated Docker container using the Docker SDK exec_start socket API for interactive GDB attachment, solving file descriptor exhaustion at the OS level. Deliverables: merged SessionManager with full test coverage, WebSocket streaming with reconnection handling, Docker sandbox with resource limits, security audit, and handover documentation.