Available immediately · London, UK

Arushi Jain

AI Engineer / Software Engineer

I build LLM-powered products end to end — from transformer NLP pipelines and RAG systems to the containerised services and production interfaces that ship them.

View my work Download CV
  • MSc Artificial Intelligence — Merit, QMUL
  • UK Graduate Visa to 2028 · open to sponsorship
Portrait of Arushi Jain
Now Lead Design Engineer @ Blocvue

01 — About

Full ML lifecycle, from notebook to production

I'm an AI and Software Engineer with an MSc in Artificial Intelligence (Merit) from Queen Mary University of London and hands-on experience building LLM-powered applications, transformer-based NLP pipelines and cloud-deployed AI services.

Most recently I've been the sole frontend engineer on a production compliance platform for UK higher-risk residential buildings, shipping 100+ routes to pilot go-live. Alongside that I'm developing an autonomous multi-agent job search platform using RAG, vector search and LangGraph.

I work comfortably across the whole lifecycle — data normalisation, model fine-tuning, containerised deployment and CI/CD with Python, GPT-4, AWS and Docker — and I care a lot about verifying what I ship rather than assuming it works.

100+ Routes shipped on a live compliance platform
92% Accuracy on an NL-to-SQL pipeline built with GPT-4 and spaCy
50% Release time cut via Dockerised AWS deploys and GitHub Actions
23 Buildings live at pilot go-live

02 — Experience

Where I've worked

  1. July 2026 — Present

    Lead Design Engineer

    Blocvue · London

    • Sole frontend engineer on a production compliance platform for UK higher-risk residential buildings, building and shipping the web application to pilot go-live across 23 buildings.
    • Designed and built the platform's design system from scratch — shared components, design tokens and a written design brief adopted as the team's specification of record.
    • Built 100+ routes in Next.js 15, React and TypeScript covering document registers, statutory obligations, risk assessments, asset registers and role-based access.
    • Led a 54-item UI audit and remediation programme, taking each issue from discovery through implementation and verification with code-level evidence.
    • Diagnosed and fixed platform-wide defects including 24 broken form navigations, unsaved-work loss, containerised session failures and cache revalidation issues.
    • Identified and escalated a database-layer authorisation gap during a permissions audit, documenting remediation before pilot data entry.
    • Next.js 15
    • React
    • TypeScript
    • Design systems
  2. April 2026

    AI Data Processing Analyst (Contract)

    Sigma.AI DATA Ltd. · London

    • Enforced annotation protocols across large-scale speech corpora for ASR and NLP model training, escalating data quality issues to pipeline leads to maintain downstream model integrity across production datasets.
    • ASR
    • NLP
    • Data quality
  3. Jan 2024 — May 2024

    Software Engineer

    StatusNeo · India

    • Engineered an end-to-end NL-to-SQL pipeline with GPT-4 and spaCy NER, letting non-technical teams query relational databases in plain English at 92% accuracy across finance and healthcare query sets.
    • Presented technical architecture and model performance findings to non-technical stakeholders, translating NLP pipeline behaviour into business-impact terms that informed product decisions.
    • Optimised a RandomForestClassifier via GridSearchCV, reducing model error by 15% and processing time by 20%; deployed on AWS via Docker with a GitHub Actions CI/CD pipeline, cutting release time by 50%.
    • GPT-4
    • spaCy
    • AWS
    • Docker
    • CI/CD

03 — Projects

Things I've built

June 2026 — July 2026

AI Job Application Copilot

An AI-powered job application platform combining LLMs, RAG, vector search and LangGraph multi-agent orchestration. Handles resume parsing, ATS optimisation, job matching, cover letter generation and live job discovery through external APIs.

A Sentence Transformers and Qdrant RAG pipeline powers semantic resume Q&A and contextual retrieval, wrapped in a human-in-the-loop lifecycle with approval workflows, status tracking and Playwright-based submission automation. Shipped full-stack on FastAPI, PostgreSQL, React and Docker Compose.

  • Python
  • FastAPI
  • OpenAI API
  • LangGraph
  • Qdrant
  • PostgreSQL
  • React
  • Docker
System flow
  1. ResumeParse + ATS
  2. RAGQdrant vector search
  3. AgentsLangGraph orchestration
  4. MatchJobs + cover letters
  5. ApproveHuman-in-the-loop
  6. SubmitPlaywright automation

Feb 2025 — Sept 2025

AI-Powered Code Assistant

  1. C codeCodeXGLUE
  2. CodeBERT+ CodeT5
  3. BiLSTMClassify
  4. GPT-4Explain

MSc thesis project. Fine-tuned CodeBERT and CodeT5 on CodeXGLUE for C code vulnerability classification, integrating a BiLSTM sequence classifier and a GPT-4 explanation layer that generates developer-readable vulnerability reports alongside predictions. All experimental parameters and precision/recall/F1 results documented for full academic replication.

  • CodeBERT
  • CodeT5
  • BiLSTM
  • GPT-4
  • PyTorch

June 2026

Spotify Reversed Playlist

A backend automation tool handling custom OAuth 2.0 logic, token refresh cycles and paginated JSON processing to intercept playlist tracks and update active playback queues in real time.

  • OAuth 2.0
  • Spotify Web API
  • Flask

Jan 2024 — May 2024

AI-Powered Social Media Automation

An automated content generation and scheduling system for LinkedIn and Instagram built on GPT-4, cutting manual creation time by 50% over a 30-day benchmarked comparison period.

  • OpenAI GPT
  • Flask
  • AWS RDS
  • Python

May 2023 — July 2023

Automated Text Structuring Pipeline

A modular NLP pipeline — tokenisation, lemmatisation, NER and relation extraction — converting unstructured documents into structured database records, a 35% improvement in downstream data usability against a manual baseline.

  • spaCy
  • NLP
  • Python
  • Flask

B.Tech thesis

Personalised Health Treatment

A multi-class symptom classification model trained on a labelled medical dataset, reaching 92% diagnostic accuracy on a held-out test set and surfacing structured precaution recommendations through a Flask web interface.

  • Scikit-learn
  • Flask
  • Python

More on github.com/arushi1106.

04 — Skills

What I work with

NLP & Transformers

  • BERT
  • CodeBERT
  • CodeT5
  • spaCy
  • NLTK
  • NER
  • POS tagging
  • Hugging Face Transformers

LLMs & Fine-tuning

  • GPT-4 API
  • OpenAI API
  • Prompt engineering
  • Text-to-SQL
  • LangChain
  • LangGraph
  • RAG pipelines

ML Engineering

  • Scikit-learn
  • TensorFlow
  • PyTorch
  • RandomForest
  • BiLSTM

Deployment & MLOps

  • Docker
  • AWS (RDS)
  • Flask
  • REST APIs
  • GitHub Actions
  • Streamlit

Data & Databases

  • Pandas
  • NumPy
  • PostgreSQL
  • MySQL
  • SQLite
  • Qdrant

Languages & Tools

  • Python
  • SQL
  • TypeScript
  • Git
  • Next.js
  • React

05 — Education

Where I studied

Sept 2025

MSc Artificial Intelligence — Merit

Queen Mary University of London · London, UK

Modules in Natural Language Processing, Machine Learning, Deep Learning, Applied Statistics and Neural Networks. Thesis on an AI-powered code assistant fine-tuning CodeBERT and CodeT5 for C code vulnerability detection.

July 2024

B.Tech Information Technology

Manipal University Jaipur · Jaipur, India

Modules in Data Structures & Algorithms, Database Management Systems, Software Engineering, Web Development and Cloud Computing. Thesis on personalised health treatment using multi-class symptom classification.

Certifications

Certified in Python, Data Science and Machine Learning.

06 — Contact

Let's build something

I'm open to AI and software engineering roles in London or remote. Send a note below — or email me directly if you prefer.

[email protected]