Hi theređź‘‹, I'm

Roshan Razak

Scarborough, United Kingdom

Technical Team Lead & Senior AI/Backend Engineer with 6+ years of experience. Specializing in shipping GenAI applications end-to-end and building the robust backend infrastructure (Python, FastAPI, AWS) that powers them.

about me.

I am a Technical Team Lead and Senior AI/Backend Engineer. My career has been defined by a natural progression from engineering rock-solid backend microservices to leading technical teams and shaping architectural vision.

experience.

  • Technical Team Lead United Kingdom

    at, dev.xabium.co.uk

    Jan 2026 – Present

    • Engineering Leadership: Stood up robust GitHub Actions pipelines and enforced rigorous PR reviews, successfully doubling deployment frequency within just eight weeks.
    • System Architecture: Architecting the multi-tenant service layer, designing schema isolation, queue workers, and robust REST API contracts to securely support downstream GenAI features.
    • Technical Strategy: Own the technical roadmap, write ADRs, and sign off on architectural trade-offs for latency vs. cost vs. model-serving flexibility.
    • Python
    • FastAPI
    • AWS
    • CI/CD

  • Software Engineer (Contract) United Kingdom

    at, cawosh.com

    Aug 2024 – Dec 2025

    • Cloud Optimization: Executed deep performance profiling of a serverless Lambda pipeline using X-Ray traces, refactoring to provisioned concurrency to cut monthly AWS infrastructure spend by 45%.
    • Data Engineering: Engineered a strict, audit-ready ETL layer using Pydantic validation to process 5,000+ documents a month, pushing clean Parquet outputs directly into the ML feature store.
    • Backend Services: Owned the Python/FastAPI services behind API Gateway, enforcing strict async/await patterns, automated IAM secrets rotation, and comprehensive pytest coverage.
    • AWS Lambda
    • Python
    • FastAPI
    • Serverless
    • ETL

  • Associate Consultant India

    at, capgemini.com

    Oct 2018 – Feb 2023

    • Deep Technical Optimization: Rewrote complex PostgreSQL queries using composite indexes, CTEs, and tuned work_mem, driving p95 retrieval times down from 12 seconds to under 2 seconds.
    • Microservices Debugging: Resolved a critical race condition in Kafka-backed Spring Boot microservices by implementing idempotency keys, eliminating silent data drift on a global banking platform.
    • Quality Assurance: Authored and enforced stringent testing standards across engineering teams, reducing production-severity bugs by 25% over 18 months.
    • PostgreSQL
    • Spring Boot
    • Kafka
    • Testing
    • Optimization

Education

Staffordshire University

Jul 2024

MSc Computer Science (Software Engineering)

Modules: Software Architecture, Distributed Systems, Cloud Computing, Security Engineering

Kannur University, India

Jul 2017

B.Tech in Computer Science and Engineering

Modules: Operating Systems, Networks, Database Systems, Software Engineering

projects.

  • DocIntel

    • An Agentic Document Intelligence Platform allowing users to fluidly chat with up to 20 complex PDFs simultaneously.
    • Powered by LangGraph agentic routing and highly-optimized hybrid search (dense + BM25 + Reciprocal Rank Fusion), enhanced by cross-encoder re-ranking for pinpoint accuracy.
    • LangGraph
    • Python
    • FastAPI
    • pgvector
    • PyMuPDF

  • InterviewPilot

    • A hyper-responsive Voice-to-Voice Mock Interviewer capable of fluid, real-time conversational assessments.
    • Engineered with FastAPI WebSockets to stream bidirectional, ultra-low-latency audio seamlessly between a React frontend and the Google Speech API (Gemini Voice).
    • Gemini Voice API
    • FastAPI
    • WebSockets
    • React
    • TypeScript

  • MemoryMesh

    • A scalable Multi-Tenant Conversational AI engine tailored for complex contextual recall over long-term interactions.
    • Features an innovative three-tier memory architecture: Redis for fast caching, pgvector for semantic similarity, and Claude Haiku for dense context summarization.
    • Redis
    • pgvector
    • Claude Haiku
    • FastAPI
    • LangChain

  • Smart Access Control

    • A Cloud-Based Facial Recognition system utilizing multi-factor authentication with hardware badge taps.
    • Built on AWS Rekognition against S3 reference sets, with DynamoDB for state management and Cognito for secure JWT issuance.
    • AWS Rekognition
    • Lambda
    • DynamoDB
    • Cognito
    • Python

  • Portfolio Website

    live preview github

    • **Serverless Deployment:** Built a fully responsive web application served cleanly from a Proxmox cluster deployment utilizing multi-stage Docker configurations natively interconnected directly over Cloudflare infrastructure.
    • **Modern UX Focus:** Prioritized readable presentation elements and modern logic layers featuring Next.js 15 capabilities, tightly interwoven with shadcn/ui components, Framer Motion aesthetics, and deeply-typed data layers.
    • Next.js 15
    • TypeScript
    • Shadcn/UI
    • Tailwind CSS
    • Framer Motion
View all projects

technical skills.

  • Core Programming:

    Python, Java, JavaScript, TypeScript, SQL, Pydantic

  • AI & Orchestration:

    LangChain, LangGraph, RAG Pipelines, Agentic Workflows, Prompt Engineering, LLM Evaluation, Ollama

  • Vector Databases:

    pgvector, ChromaDB, Embedding Models, Semantic Search

  • Backend & Web:

    Next.js, ReactJS, FastAPI, Spring Boot, Laravel, Django, shadcn/ui, REST APIs

  • Databases:

    PostgreSQL, Prisma, MySQL, Redis, DynamoDB

  • DevOps & Infrastructure:

    Docker, Kubernetes, Terraform, Jenkins, GitHub Actions, AWS Lambda, S3, EC2, Bedrock, CI/CD