Building intelligence by design.
I architect AI systems that think, learn, and adapt — from fine-tuned LLMs to real-time computer vision pipelines.
I make machines that understand the world.
5 years building ML systems at Google DeepMind, Stripe and beyond. My obsession: making AI feel less like a tool and more like a collaborator.
- 01Large Language Models
- 02Computer Vision
- 03Generative AI
- 04MLOps & Infrastructure
- 05Reinforcement Learning
Building RAG pipelines serving
500K+ daily users.
Projects that
ship and scale.
Enterprise RAG Chatbot
A production-grade Retrieval-Augmented Generation system powering internal knowledge search for enterprise teams. GPT-4 + Pinecone vector DB, serving 500K+ queries daily with sub-200ms latency.
Real-Time Object Detection API
Custom-trained YOLOv8 model hitting 60 FPS on edge hardware. Zero-latency REST API with batch inference support — used in smart retail and industrial safety monitoring at scale.
Financial Sentiment Engine
Fine-tuned BERT model on 10M+ financial news articles, achieving 94% classification accuracy. Powers live trading signals at a hedge fund — 35% increase in signal confidence versus baseline.
AI Art Style Transfer Studio
Stable Diffusion fine-tuned with custom LoRA adapters trained on curated artist styles. Gradio web interface lets creators apply styles instantly — deployed on AWS with auto-scaling GPU pods.
AutoML Pipeline Platform
End-to-end MLOps system: automated data ingestion, feature engineering, hyperparameter tuning with Optuna, experiment tracking via MLflow, and one-click Kubernetes deployment with Terraform.
Tools I actually use.
Where I've
made impact.
Senior AI Engineer
OpenAI PartnersLead 6-person team building LLM products. RAG pipelines for 500K+ users, 60% latency reduction via quantization.
ML Engineer
Google DeepMindProductionized RL agents on TPU clusters. Processed petabytes of training data. Co-authored 2 NeurIPS papers.
Data Scientist
StripeFraud detection models cutting chargebacks by 35%. Real-time scoring system handling 1M+ transactions per day.
AI Research Intern
Stanford AI LabResearched novel attention mechanisms for vision-language models and benchmarked SOTA approaches.
What I'm expert at.
Large Language Models
Fine-tuning, prompt engineering, RAG pipelines, and production deployment of LLMs including GPT-4, LLaMA, Mistral, and custom models.
Computer Vision
Object detection, segmentation, real-time inference at the edge. Specializing in custom YOLO, DETR, and diffusion-based vision systems.
MLOps & Infrastructure
End-to-end ML pipelines, experiment tracking, automated retraining, and one-click deployment on Kubernetes & cloud platforms.
Generative AI & Multimodal
Text-to-image, video synthesis, multimodal agents. Stable Diffusion, LoRA fine-tuning, and CLIP-based retrieval at scale.
NLP & Conversational AI
Sentiment analysis, entity extraction, summarization, and dialogue systems using BERT, T5, and custom transformer architectures.
Real-Time AI Systems
Low-latency AI inference for production use cases — edge deployment, model quantization, TensorRT optimization, and streaming pipelines.
Validated expertise.
Credentials from leading industry organisations, validating hands-on skills in AI, ML, and cloud platforms.
Professional Machine Learning Engineer
Demonstrates the ability to design, build, and productionize ML models using Google Cloud — Vertex AI, BigQuery ML, and TFX pipelines serving at scale.
AWS Certified Machine Learning – Specialty
Validates expertise in building, training, tuning, and deploying ML models using the AWS cloud — SageMaker, data engineering, and model monitoring.
Deep Learning Specialization
5-course program covering neural networks, hyperparameter tuning, CNNs, sequence models, and structuring production ML projects — by Andrew Ng.
Natural Language Processing with Transformers
Hands-on covering BERT, GPT, T5, and modern NLP pipelines using the Hugging Face ecosystem — from fine-tuning to deploying in production.
PyTorch for Deep Learning & Neural Networks
End-to-end mastery of PyTorch — tensors, autograd, custom architectures, distributed training, and model export with TorchScript and ONNX.