Postgraduate
University of Birmingham, Dubai
MSc Artificial Intelligence and Machine Learning. Completed September 2026.
I build AI systems that work in production.
AWS Certified Solutions Architect with an MSc in AI and Machine Learning, based in Dubai. I work on LLM powered agentic workflows, RAG architectures, and cloud native AI infrastructure.
Certified by
01 — About
Cloud infrastructure, and the models that run on it.
I have an MSc in Artificial Intelligence and Machine Learning from the University of Birmingham Dubai, and I am an AWS Certified Solutions Architect. I work where cloud infrastructure meets machine learning, mostly on agentic workflows, retrieval systems, and the infrastructure they run on.
As a Cloud Engineer intern I designed and built a global serverless platform on AWS, tuning content delivery with CloudFront and ElastiCache and setting up a pilot light disaster recovery strategy to protect availability. That is where I learned to turn a list of technical requirements into something that actually runs.
I am open to Dubai, Abu Dhabi or Riyadh based AI and cloud native roles, or remote.
- BE, CGPA
- 8.79
- Based in
- Dubai
02 — 3 roles
Experience
01
December 2025 to September 2026
Remote
AI Solutions Consultant
HCF Tech Services
Led the design and development of AI driven document intelligence solutions using cloud native architectures, OCR and LLMs, bridging business requirements and technical implementation to ship enterprise focused AI products.
The main piece of that work is BidInspect, an AI assisted tender evaluation workbench for procurement committees. It creates the tender, takes the bidder documents, extracts the requirements and analyses the evidence, then hands the committee a verification pass over anything it is not confident about before mapping criteria and scoring. Supporting evidence stays attached to each judgement and compliance gaps get surfaced, while the decision itself stays with the constituted committee.
- Four stages: criteria, submissions, verify, results
- Committee ready outputs, a Comparison Brief and a Comparison Matrix
- Document integrity and evidence accuracy checks before scoring
- AI Solution Architecture
- LLM evaluation
- Document intelligence
- RAG
- OCR
BidInspect, tender evaluation workbench
02
June to September 2025
Pune, hybrid
Cloud Engineer, internship
Verve Communications
Managed a programme to improve a global content delivery platform by architecting a scalable serverless AWS solution to cut latency. Built a multi layered caching strategy to lower cost and server load, designed a pilot light disaster recovery strategy for business continuity, and defined a CloudWatch monitoring framework.
- AWS
- Serverless
- CloudFront
- ElastiCache
- CloudWatch
Control plane architecture
03
January to March 2025
Pune, on site
AI Intern
AI & DS Department, PVG's COET & GKPIOM, Pune
Led development of ZenKraft, a full stack personalised yoga training system doing real time pose estimation and corrective feedback. Integrated the ML models, applied OpenCV for image processing to reach 92% accuracy on posture corrections, built a 30 posture dataset with experts, and wrote the user facing application that surfaces insights and performance metrics.
- Presented at the BC2AD 2025 conference in Pune
- Published in JOSH-PE, Journal of Sports Health and Physical Education
- CNN
- OpenCV
- Mediapipe
- RAG
- Python
ZenKraft system architecture
03 — Selected work
Projects
AEGIS
LLM security engine. Open source, live demo on Hugging Face.
Traditional security tools look for keywords, but attackers use context, roleplay and social engineering to get past them. AEGIS runs a four layer defence grid instead: a reflex tier of regex and heuristics that blocks script attacks at zero latency, a sentinel layer running Llama 3 on Groq that analyses intent in real time, a self healing vector memory, and a context aware auditor that uses RAG to enforce company policy on the output.
When it detects a zero day, it immunises its vector memory and fails over to a honeypot model, so the same attack does not work again. The core engine is open source and designed to be containerised and deployed on premise for data sovereignty compliance.
- Added latency
- <200ms
- Defence layers
- 4
- Intent check
- <300ms
- Licence
- Open
- Llama 3
- Groq LPU
- ChromaDB
- LangChain
- Gradio
- Docker
Live demo on Hugging Face
Source
Serverless photo sharing
AWS, fully serverless
A fully serverless web app on AWS. Guests scan a QR code at an event and upload photos straight from their phone, into one central bucket.
- S3, Lambda, API Gateway, DynamoDB and IAM
- API Gateway fronts the Python Lambda functions
- DynamoDB tracks photo counts per device per event
- AWS Lambda
- API Gateway
- S3
- DynamoDB
- IAM
Forex AI trading bot
Reinforcement learning
A reinforcement learning agent trained to automate trading strategies, in a custom simulation environment with risk and reward criteria and penalties.
- DQN and PPO agents on TensorFlow with CUDA
- Trained on over 200,000 data points
- Best model returns 52% across six months of data
- Reinforcement Learning
- TensorFlow
- Keras
- Pandas
- CUDA
04 — Newest first
Activity
Drag to move through it
August 2026
BidInspect product demo
Published a full walkthrough of the tender evaluation workbench I have been building at HCF Tech Services, run end to end on a synthetic procurement.
Anthropic Claude Academy, two tracks
Completed Building with the Claude API and Model Context Protocol: Advanced Topics. Standardised protocol layers are becoming the plumbing for production LLM architecture.
AEGIS, an immune system for LLMs
Open sourced a four layer defence engine that learns from the attacks it blocks, and put a live demo on Hugging Face for people to try to break.
NVIDIA DLI, Generative AI with Diffusion Models
Built and deployed diffusion models to generate high quality images, which filled in how modern generative architectures actually work under the hood.
MSc begins
Started the MSc in AI and Machine Learning at the University of Birmingham, Dubai.
AWS Certified Solutions Architect, Associate
Passed the SAA-C03 exam. Cloud architecture is where creativity meets engineering, and it is the backbone I want under the AI work.
Paper published
"Integrating Artificial Intelligence with Yoga: A Systematic Literature Review for ZenKraft" appeared in JOSH-PE, Volume 5, Special Issue.
Presented at BC2AD 2025, Pune
Our team presented ZenKraft at the conference. The feedback from the panel changed how we framed the research.
05 — Certifications and a paper
Credentials
Amazon Web Services
Certified Solutions Architect, Associate
Verify
NVIDIA
Generative AI with Diffusion Models
Verify
Anthropic
Building with the Claude API
Verify
Anthropic
Model Context Protocol, Advanced Topics
Verify
NVIDIA
Fundamentals of Deep Learning
Done
Udemy
PyTorch for Deep Learning
Done
Publication, April 2025
Integrating Artificial Intelligence with Yoga: A Systematic Literature Review for ZenKraftJOSH-PE, Journal of Sports Health and Physical Education. Volume 5, Special Issue.
Published
06 — Tools I use
Skills
AI and ML
Large Language Models, RAG, Agentic AI, LangChain, TensorFlow, PyTorch, Scikit-learn, OpenCV, Mediapipe, Pandas, NumPy
AWS and cloud
S3, EC2, Lambda, API Gateway, DynamoDB, RDS, Route 53, CloudFront, ElastiCache, IAM, Docker
Languages
Python, JavaScript, SQL, C++, HTML and CSS
07 — Contact
Open to Dubai, Abu Dhabi or Riyadh based AI/Cloud native roles, or remote.
If you have a question or want to work together, the fastest route is email.