Building Reliable ML systems that ship.
I design, train, and ship production ML systems — from time-series forecasting and NLP to real-time model serving.
Featured Work
Live Gold Price Forecaster
Developed a Transformer ensemble forecasting XAUUSD. Ships live signals via WebSocket.
Production NLP Pipeline
End-to-end news ingestion → classification → sentiment drift monitoring with prompt + finetuned models, 10k docs/hr.
Computer Vision Inspection
Lightweight CNN + ONNX Edge runtime catching defects on assembly line cameras at 30 FPS on Jetson.
ML Platform & Feature Store
PostgreSQL feature store, Airflow pipelines, FastAPI model gateway, and full-stack ML observability.
Time-Series Backtesting
Walk-forward validation, slippage, equity/drawdown analytics, and promotion to live.
LLaMA-8B News Sentiment
LoRA-tuned 8B model for finance news sentiment with uncertainty routing. And got a 99.9% Accuracy.
- AWS ML Specialty2024
- deeplearning.ai MLOps2023
- Built end-to-end ML pipelines with monitoring, CI/CD, and human-in-the-loop feedback.
Skills & Tools
Modeling
- Time-Series (XGBoost/Transformers)
- NLP (LLMs, RAG)
- CV (CNN/ONNX)
- Optimization
MLOps
- Experiment Tracking
- Feature Stores
- CI/CD for ML
- Observability & Drift
Platform
- FastAPI/GRPC
- Airflow/K8s
- PostgreSQL/BigQuery
- Kafka/Lambda
Publications & Writing
About
I’m a hands-on ML engineer who loves the messy middle: turning noisy data into deployable models and user-visible impact. I care about robust baselines, clean data contracts, and shipping small wins fast.
Let’s build something
Open to roles and consulting. I’m especially excited about time-series forecasting, production NLP, and ML platform work.