Mohammed Affan
Software Developer · Known online as Itachi
Professional profile
Software developer working across AI systems and the infrastructure around them. At Pollinations.ai, contributes agent interfaces, MCP tooling, retrieval, and isolated execution; at Myceli.ai, builds AI workflows and automation. Independent work spans model research, Rust systems, evaluation environments, and web, mobile, and 3D applications.
Professional experience
Pollinations.ai
Feb 2025 – PresentDeveloper · MLOps & AI Infrastructure
- Build AI agents, model integrations, APIs, and workflow automation for the Pollinations platform.
- Extended the MCP server with tools for image batches, video generation, multimodal analysis, transcription, search, and pricing; added dynamic model discovery and API-key authentication.
- Contribute to Polli across Discord, GitHub, and OpenAI-compatible APIs, connecting repository search, session context, caller-scoped tools, model routing, and streamed responses.
- Moved repository embeddings into Cloudflare Vectorize with incremental indexing, batched embedding requests, and removal of vectors for deleted files.
- Replaced per-user database update subprocesses with batched operations and concurrent API validation; isolated Cloudflare shell execution in disposable containers without caller credentials.
Myceli.ai
Feb 2025 – PresentDeveloper · AI Systems & Automation
- Connect AI infrastructure through automations and workflows, connecting models, services, and agent modules across cloud and local environments.
Independent research & development
Ouroboros
2026 · Independent researcher & developer · Research in development
- Develop language-model research prototypes in PyTorch under local hardware constraints; retain unpublished architecture and implementation details privately.
- Evaluate mechanisms independently against explicit baselines; document ablations, limitations, and negative findings in a public research notebook.
- Public output is a research log. No published paper or public benchmark result is claimed.
Sylph
2026 · Independent systems developer · Private implementation in development
- Develop a Rust-based AI runtime exploring local inference, task-driven learning, and peer-to-peer exchange of model adapters.
- Separate model backends through an inference contract; implement adapter compatibility and payload-integrity checks in the exchange path.
- Develop adapter-composition checks and adversarial tests, with local and Kubernetes/Helm deployment tooling. Public overview available; implementation remains private.
Selected engineering projects
Memory-Arc
2025 · Independent developer · Public Python implementation
- Built configurable memory infrastructure connecting recent conversation history to persistent vector retrieval, with AI, heuristic, hybrid, and disabled processing modes.
- Separated configuration, JSON persistence, asynchronous memory interfaces, and model-provider adapters.
Repository intelligence
2026 · Independent developer · Private Rust tooling in development
- Develop local repository-analysis tooling using Rust, Tree-sitter, SQLite, and FTS5 for language structure, persistence, and evidence retrieval.
- Build explicit validation context to distinguish stored results from evidence about the current source state.
Agent evaluation environments
2026 · Environment developer · Public OpenEnv hackathon projects
- Built simulation environments for evidence-based investigation, with procedural scenarios, MCP tools, state transitions, and reward components.
- Added adversarial tests for reward shortcuts and evidence gathering; developed an adversary that advances during investigation and adapts to containment actions.
Application engineering
2026 · Application developer · Private web, mobile & 3D implementations
- Built a Flutter records and reporting application with Supabase integration, insights screens, PDF tooling, and automated builds.
- Developed a Next.js presentation builder with research, outline, image-generation routes, streamed results, and editing interfaces.
- Built SceneForge with typed scene descriptions, incremental JSON parsing, and streamed updates to an interactive Three.js scene.
Technical skills
Languages: Python, TypeScript / JavaScript, Rust, Dart, SQL.
AI & MLOps: PyTorch, model fine-tuning, local inference, model deployment, MCP, agent orchestration, RAG.
Infrastructure: Cloudflare Workers, Vectorize, D1, Docker, Kubernetes, Helm, Linux, GitHub Actions.
Applications & data: FastAPI, React, Next.js, Flutter, Three.js, Qdrant, SQLite, Supabase.
Education
Diploma — AI/ML pathway
Pursuing · 5th semesterMS Ramaiah Polytechnic
Selected contributions · Pollinations.ai
Indexing · PR #12602 — Changed paths are reindexed; deleted paths have their vectors removed.
Isolation · PR #14787 — Shell work runs behind an explicit container boundary.
Streaming · PR #14438 — Clients can receive incremental output through a familiar API interface.