Jazil Imran

BA Computer Science. Courses include Algorithms, Object-Oriented Programming, Operating Systems, Databases and Software Engineering.

A-levels: A*A*A*A* in Maths, Further Maths, Computer Science and Physics. Academic Achievement Prize, Computer Science Prize, Senior Maths Challenge Gold, Cricket Team Captain.

• Designed and delivered lessons, presentations and Packet Tracer labs for 20+ CCNA students.
• Taught IPv4, subnetting, VLANs, routing protocols, DNS and NAT, plus troubleshooting with ping, traceroute and show commands.
• Covered security: ACLs, LAN port security, encryption and firewall fundamentals.

• Shadowed software and data engineering teams building object detection, performance analysis and video processing at scale.
• Followed the delivery workflow across version control, code review, testing, containerisation and CI/CD.

• macOS menu-bar agent: hold ⌘⇧Space, speak, and it acts in the frontmost app through the accessibility tree.
• Fills job applications completely in page order, drafting open answers from the user’s CV with Gemini.
• Every click and keystroke is checked by a pure, tested policy function; it never submits, sends or enters credentials.

• Built an autonomous agent that detects and contains app failures to stabilise services before engineers intervene.
• Orchestrated concurrent repair races across isolated Modal sandboxes, one per Gemini-generated fix, with Jev and a computer-use agent running real browser journeys and failure-injection tests.
• Validated Gemini outputs with strict Pydantic schemas, ranked patches through deterministic test gates, and activated the smallest safe patch under a reversible lease.

• Real-time encrypted messaging, group chat and video calling for users anywhere on the internet.
• End-to-end encryption with X25519 key exchange for direct messages and rotating AES-256-GCM keys for groups.
• Zero-knowledge authentication where passwords never leave the device, using Argon2id, JWT rotation and one-time codes.
• Live presence and typing over WebSockets; peer-to-peer WebRTC calls with TURN relays, containerised with end-to-end tests.

• Full-stack AI revision platform for Cambridge students: 230+ past questions, marking and a weakness dashboard.
• LLM marking pipeline with model routing and rubric generation grounded in official mark schemes, validated by a pytest eval harness.
• Secured against prompt injection with structured-output validation and output sanitisation.
• Containerised React + FastAPI app deployed to Render via GitHub Actions, with Gemini and Google OAuth.

• Trained pose-based classifiers for seven batting shots and six bowling actions with MediaPipe, evaluated leave-one-session-out.
• React + FastAPI service that turns uploaded footage into pose-tracked replays, biomechanics metrics and evidence-cited coaching.
• GitHub Actions run backend, frontend and Docker health checks on every change.

• Interactive simulator comparing link-state (Dijkstra) and distance-vector (Bellman-Ford) routing on editable topologies.
• Custom binary min-heap with O(log n) decrease_key; live forwarding tables, packet tracing and a count-to-infinity demonstration.