Air Quality Forecasting Platform
forecasts the next 24 hours from live government data every morning, then scores itself against what actually happened.
- Python
- XGBoost
- Evidently
- FastAPI, GitHub Actions, Hugging Face
machine learning engineer.
lately, computer vision on a production line.
(more interested in how models behave in production than how they're trained.)
01 · work
what does a model need to survive contact with real data?
featured
more
Air Quality Forecasting Platform
forecasts the next 24 hours from live government data every morning, then scores itself against what actually happened.
- Python
- XGBoost
- Evidently
- FastAPI, GitHub Actions, Hugging Face
Emotion-Aware Insight Generator
reads the emotion on a face, then finds and summarizes the reviews that explain it.
- ResNet-50
- FAISS
- LangChain
- Flan-T5, RoBERTa, Streamlit
Brain Tumor MRI Classifier
97.15% test accuracy, after augmentation first made it worse.
- TensorFlow
- Keras
- CNN
- Flask
PDF Question Answering with RAG
answers questions about an uploaded PDF from the passages it retrieves, not from memory.
- LangChain
- ChromaDB
- MiniLM
- Gemini, Flask
Electricity Load Forecasting (Time Series)
forecasts a full day of 15-minute load (96 steps), and rebuilds a research algorithm as an R package.
- scikit-learn
- XGBoost
- SARIMA
- LSTM, R
team project
Machine Learning for Diabetes Risk Prediction
96% accuracy, 0.993 AUC, and a web form to act on it.
- XGBoost
- Random Forest
- FastAPI
02 · experience
where i've worked
software engineering and data science roles, contributing to ML workflows, automation, and production inference.
Data Science Intern
STMicroelectronics
Rousset, France
Part of the Quality, Manufacturing & Technology (QMT) – Data Science team. Improved production-line efficiency by developing a deep learning object detection solution that replaces manual operator inspection of solder-ball defects in semiconductor packages.
- Used SAM3 segmentation to automatically crop the region of interest from raw X-ray images, and generated training labels with a grid-based annotation method.
- Fine-tuned a YOLO11 object detection model to locate defective solder balls, raising accuracy from 70% to 90%, with experiments tracked in MLflow.
- Developed and deployed a modular Python pipeline that automatically processes new images as they arrive and outputs annotated images with the precise location of each defect. Loading all models only once for the entire pipeline cut processing time from 2 seconds per image to milliseconds.
- Pushed inference results to a MariaDB database, with processed images stored on a network storage server (NAS).
- Built a FastAPI backend and a Vue.js web dashboard that display each inspection result in real time.
- Collaborated with data engineering and manufacturing teams to align the solution with production requirements.
- Python
- SAM3
- YOLO11
- MLflow
- FastAPI
- Vue.js
- MariaDB
Sr. Software Engineer
Capgemini Technologies Pvt Ltd
Hyderabad, India
Supported ML-driven and data-intensive features within production-grade systems.
- Supported data pipelines used for analytics and machine learning workflows.
- Collaborated with data scientists to operationalize preprocessing and feature generation logic.
- Worked on containerized services and CI/CD pipelines to enable consistent and reliable deployments.
- Assisted with monitoring, logging, and debugging of data-driven services in production.
Data Science Intern
Digital Lync
Hyderabad, India
Supported end-to-end ML experiments from datasets to model evaluation.
- Prepared datasets through cleaning, normalization, and feature extraction.
- Trained and evaluated classical ML models using metrics such as accuracy, precision, recall, and AUC.
- Performed exploratory data analysis and error analysis to guide model improvements.
- Documented experiments and results to support iterative model development.
03 · education
Master's in Data Science & Artificial Intelligence
- Advanced coursework in machine learning, deep learning, data engineering and MLOps.
- Hands-on projects in forecasting, computer vision and LLM/RAG-based applications.
- Focus on designing end-to-end ML workflows, from experimentation to reliable delivery.
Bachelor's in Electronics and Communication Engineering
- Built strong foundations in mathematics, signal processing and communication systems.
- Developed core programming, problem-solving and engineering skills.
- Academic projects that sparked interest in data-driven systems and ML.
04 · hello
i build things that have to survive the real world, not just the demo.
about
a machine learning engineer and data scientist working across computer vision, NLP and LLMs/RAG.
i hold a master's in data science & AI (2026), with a focus on rigorous experimentation, model evaluation, and production-grade MLOps pipelines that scale.
most recently, i took an AI-based X-ray inspection system at STMicroelectronics from prototype to production deployment, fine-tuning its YOLO model from 70% to 90% accuracy.
skills
- Python
- SQL
- OpenCV
- YOLO
- SAM3
- PyTorch
- CNN
- ResNet
- Hugging Face
- FastAPI
- Flask
- WebSockets
- NumPy
- scikit-learn
- XGBoost
- MariaDB
- ChromaDB
- MLflow
- DVC
05 · contact
let's connect.
open to new opportunities and collaborations, reach out below.
say hi
send me the interesting problem. machine learning, computer vision, mlops, anything worth building.
elsewhere