I build robust, scalable artificial intelligence and machine learning pipelines. With deep expertise in deep learning, LLMs, and computer vision, I architect intelligent solutions that enterprises trust.
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
class ModelInference:
def __init__(self, model_name: str):
self.tokenizer = AutoTokenizer.from_pretrained(model_name)
self.model = AutoModelForCausalLM.from_pretrained(
model_name, torch_dtype=torch.float16, device_map="auto"
)
def generate(self, prompt: str) -> str:
inputs = self.tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = self.model.generate(**inputs, max_new_tokens=100)
return self.tokenizer.decode(outputs[0])I am an AI/ML engineer with a strong foundation in deep learning and scalable model deployment. My work revolves around designing intelligent architectures, from deploying state-of-the-art LLMs to building robust data pipelines. I have a deep passion for natural language processing and Computer Vision.
Having navigated the complexities of high-scale systems in predictive analytics, Generative AI, and MLOps domains, I understand that clean engineering isn't just an aesthetic choice—it's a critical component of model performance and reliability.
PyTorch, TensorFlow, Scikit-Learn, Deep Learning.
Transformers, LangChain, RAG, Prompt Engineering.
AWS SageMaker, MLflow, Docker, Kubernetes.
Pandas, Spark, SQL, Vector Databases.
Solving complex business problems with cutting-edge ML models.
Optimizing weights, quantization, and accelerating serving with GPUs.
From distributed training clusters to seamless model serving architectures.
A highly scalable Retrieval-Augmented Generation (RAG) platform delivering precise answers from millions of internal documents. Employs advanced semantic search and integrates with state-of-the-art open-source LLMs.
Real-time object detection and semantic segmentation gateway for autonomous systems. Processes high-definition video streams with sub-millisecond overhead per frame.
Advanced time-series forecasting system to predict logistics supply chain bottlenecks up to two weeks in advance. Built using gradient-boosted trees and deep learning models.
Custom fine-tuned stable diffusion model that generates brand-compliant marketing assets efficiently and safely, leveraging custom LoRA weights.
"One of the strongest AI engineers I've worked with. Their ability to distill complex research papers into production-ready PyTorch implementations is unmatched."
"Pioneered our shift to generative AI. They don't just train models; they architect AI systems that are scalable, reliable, and exceptionally accurate."
Currently open to new opportunities, specialized freelance projects, and collaborations.