RS

AI Engineer in London

Rokesh Sankar

I build production-grade AI systems across NLP, RAG, LangGraph workflows, evaluation, and cloud deployment.

See work Download CV
Portrait of Rokesh Sankar
Available for AI Engineer roles Full right to work in the UK
RAG LangGraph LangChain PyTorch Model evaluation MLOps AWS Azure GCP Docker

MSc Artificial Intelligence with Distinction from Queen Mary University of London. Two-plus years building AI-first software, automation, data pipelines, and applied NLP systems for enterprise and research contexts.

Selected work

Useful AI systems, from search to software understanding.

Open source

LegacyDecoder

Turns undocumented legacy code into living documentation using AST parsing, SIE analysis, and Notion publishing.

  • HTML
  • AST parsing
  • Documentation systems
View repository

Research

Machine-Generated Text Detection

Fine-tuned RoBERTa and DeBERTa-v3 to detect AI-written text and author boundaries, reaching 84% accuracy and 0.86 F1.

  • NLP
  • Transformers
  • PyTorch

Information retrieval

Legal Search Engine

Built retrieval over 10k+ legal documents with TF-IDF and BM25, improving search precision by 22%.

  • BM25
  • TF-IDF
  • Python

Practice

LeetCode 75

Python solutions for core data structures and algorithms, kept as a public record of steady problem-solving practice.

  • Python
  • Algorithms
  • DSA
View repository

Story

I like the messy middle where research becomes reliable software.

My work sits between machine learning and engineering: building pipelines, testing ideas, measuring models, and turning prototypes into things teams can run.

At Cognizant, I worked across AI maturity modelling, Python automation, metadata ingestion on AWS, and healthcare data pipelines with Azure Data Factory and Databricks.

At Queen Mary, I focused deeper on NLP, information retrieval, neural networks, ethics, and the practical details of evaluating AI systems under pressure.

2+ years
AI-first engineering experience
Distinction
MSc Artificial Intelligence
30+ clients
AI maturity framework impact

How I work

Clear experiments, measured outcomes, maintainable systems.

01

Frame the task

Define the user goal, data shape, risks, and evaluation criteria before chasing a model choice.

02

Build the loop

Prototype retrieval, prompts, embeddings, fine-tuning, or agents with observability from the start.

03

Ship responsibly

Automate tests, document tradeoffs, and keep deployment simple enough for teams to maintain.

Connect with me