London, UK
Hi, I'm Orcas.
I build practical software and AI-powered products — recently focusing on an AI Shopping Assistant (NLP, ranking, search) and scalable web backends.
Currently
Open to Graduate / Early Career roles
Interested in Software Engineering, Applied ML, and data-driven product experiences — especially where AI meets search, ranking, and real user journeys.
Focus
Full-stack + Applied AI
End-to-end building: data → logic → UI → deploy.
Strengths
Python · TypeScript · SQL
Clean architecture, readable code, measurable outcomes.
Goal
UK roles (London-friendly)
Graduate / Early Career — open to hybrid teams.
London, UK
About
I'm Orcas, a Computer Science graduate in the UK. I build user-focused software and applied AI projects — especially systems that connect NLP to real product experiences (search, ranking, and recommendations).
What I'm looking for
Graduate / Early Career roles in Software Engineering, Applied ML, or Data-driven Product — where I can ship production-quality features, work with real data, and keep building end-to-end systems.
Strength
End-to-end ownership
From data + logic to UI and deployment.
Interest
NLP + Search/Ranking
Practical applied AI with measurable outcomes.
Location
UK (London-friendly)
Open to hybrid teams.
Highlights
- Built an AI Shopping Assistant with intent classification, attribute extraction, fuzzy search, and ranked recommendations.
- Comfortable across backend logic, data pipelines and lightweight UIs to deliver usable products.
- Strong foundation in Python, JavaScript/TypeScript, SQL and practical experimentation.
- Awarded 2nd place at a hackathon for Mira — a voice AI companion that generates a personalised 紫微斗数 chart from birth data and opens with a spoken introduction in character, built end-to-end in 24 hours.
Core skills
Languages
Python, JavaScript/TypeScript, SQL (plus Java fundamentals)
AI / Data
Transformers/NLP, classification, data cleaning, evaluation, experiment mindset
Engineering
APIs, modular architecture, Git, deployable web apps (Next.js), cloud basics (Workers/Vercel)
Experience
Bank of China (UK) — Technology Intern
Internship
- Supported reporting and operational workflows using structured data and automation-focused thinking.
- Strengthened attention to detail, communication, and working with stakeholders under deadlines.
Mira — Hackathon, 2nd Place
24-hour Sprint
- Built a voice-first AI companion that parses a 紫微斗数 (Zi Wei Dou Shu) birth chart using iztro, maps the user to one of four archetypal personas, and delivers a spoken introduction personalised to their chart — all with zero onboarding friction.
- Integrated Alibaba DashScope (Qwen) for LLM inference and GLM-4-Voice for real-time spoken dialogue; shipped a full Next.js 14 app — onboarding → animation → voice encounter — within 24 hours.
Independent Projects
Portfolio
I actively build and iterate on practical projects that combine AI + software engineering — with clean READMEs, reproducible setup, and deployable demos where possible.
Education
- MSc Computer Science (2:1) — Queen Mary University of London
- BSc Computer Science (2:1) — University of Birmingham
How I work
I like building things that are easy to run, easy to review, and easy to extend. I prioritise clarity: clean architecture, readable documentation, and measurable outcomes.
Selected work
Projects
A focused selection of work demonstrating practical engineering and applied AI — with clear outcomes, clean architecture, and deployability.
Showing 5 projects.
AI Shopping Assistant
FeaturedNLP-driven product recommendation for fashion queries
An end-to-end AI shopping assistant that understands natural language shopping requests and returns ranked product recommendations. Built with modular NLP components and a structured product database to support experimentation and future model comparisons.
Highlights
- Designed an NLP pipeline: intent classification, style extraction, and price parsing
- Implemented fuzzy matching + ranking over a structured product database (e.g., MySQL)
- Built a clean, modular architecture to support ablation studies and model/API comparison
Tech
Cloudflare AI Agent (Workers + Durable Objects)
FeaturedEdge-deployed AI agent with memory and prompt documentation
A lightweight AI agent web app deployed on Cloudflare Workers, using Durable Objects for stateful memory. Designed for simple UX, clear prompts, and reproducible setup — ideal as a practical systems + AI integration demo.
Highlights
- Implemented conversational memory using Durable Objects (stateful edge storage)
- Built a clean UI and deployment workflow for fast iteration
- Documented AI prompts and system behaviour for transparency and reproducibility
Tech
Mira — AI Companion (Hackathon 2nd Place)
FeaturedZero cold-start AI companion powered by 紫微斗数 astrology + real-time voice
Built for an AI Aesthetics 24-hour hackathon — awarded 2nd place. Mira is a voice-first AI companion that knows you from the moment you arrive. Enter your birth date and time, and Mira generates a personalised Zi Wei Dou Shu (紫微斗数) chart, maps you to one of four archetypal personas, and opens with a spoken introduction drawn from real chart insights. Entire codebase written with Claude as the AI coding environment.
Highlights
- Designed a full 紫微斗数 pipeline using iztro — parses birth data into 12-palace charts and extracts dominant stars across Self, Relationships, and Career palaces
- Engineered four distinct LLM personas (Sovereign, Flame, Seer, Warmth) with bilingual (EN/CN) system prompts derived directly from chart output
- Integrated Alibaba DashScope (Qwen) for LLM inference and GLM-4-Voice for real-time spoken dialogue; shipped the full app within 24 hours using Claude as the primary coding environment
Tech
BookmarkHub
FeaturedFull-stack bookmark manager with auth, tags, and public sharing
A full-stack bookmark manager where users can save, organise, and share links. Features JWT authentication, tag-based filtering, search, and a public feed — built end-to-end with a clean REST API and a responsive React UI.
Highlights
- Built a REST API with Express and Prisma ORM — register, login (JWT), and full bookmark CRUD
- Implemented tag filtering, full-text search, and public/private visibility per bookmark
- Designed a React frontend with protected routes, modal forms, and a toggle-based public feed
Tech
Personal Portfolio Website
Minimal, recruiter-friendly portfolio with fast performance
A clean personal website to showcase projects, experience, and contact details. Focused on strong information hierarchy, readability, and deployability.
Highlights
- Built with Next.js App Router + Tailwind for a clean, responsive layout
- Deployed to Vercel with automatic updates on push
- Structured pages for fast scanning: Home, Projects, About/CV, Contact
Tech
Let's connect
Contact
If you're hiring for Graduate / Early Career roles, I'd love to chat. Choose the channel that's easiest for you.
Orcas
London, UK
Best for applications and quick reach-outs.
orcasjiovo@gmail.com
Click to open →
Connect for roles, referrals, and updates.
linkedin.com/in/orcas-ji-about-me
Click to open →
GitHub
Browse my projects, code, and READMEs.
github.com/OrcasJi
Click to open →
Quick note
For applications, email is usually fastest. For networking and referrals, LinkedIn works best.