Data Engineering · Machine Learning · MLOps
We design, build, and operate production-grade AI systems, data pipelines, and ML infrastructure — so you can focus on your product.
What We Do
Scalable ETL/ELT pipelines, data lake architectures, and real-time streaming. We work with Spark, Snowflake, Airflow, AWS Glue, and more to turn raw data into reliable, query-ready assets.
End-to-end ML solutions — from LLM-powered chatbots and RAG systems to computer vision and NLP. We build, fine-tune, and deploy models that deliver measurable business value.
Production ML infrastructure — CI/CD for models, automated training pipelines, monitoring, autoscaling, and cost optimization. We make your ML systems reliable, fast, and affordable.
Portfolio
A selection of projects we've delivered across Data Engineering, Machine Learning, and MLOps.
AIIME
Production AWS infrastructure for an AI personal assistant on the App Store. End-to-end MLOps pipeline with cost-optimized RAG chatbot, vector DB integration, and LLM orchestration.
Stealth Startup
AI assistant providing legal answers with relevant sources (laws, court cases, opinions) using RAG with Pinecone and LLMs. Achieved over 90% answer accuracy.
Stealth Startup
Complex AWS infrastructure with multiple AI agents and data pipelines. GNNs, LLMs, NLP models deployed to EKS with SageMaker pipelines for graph data processing.
Stealth Startup
AI-powered tutor that adapts pace and content format (image, video, game, voice) to student understanding level. Audio mode for voice-based learning interaction.
Provectus
Enterprise document summarization platform using LLMs, RAG, and NLP. Led a team of 8 engineers. Summarized 200+ page insurance reports at $1/document with high precision.
Stealth Startup
AI platform transforming raw content (video, audio, PDFs) into interactive step-by-step guides. Medallion data architecture with multimodal LLM processing and context-aware AI assistant.
Financial Times
Model deployment through SageMaker and CloudFormation. Autoscaling, cost optimizations, model optimization, and FastAPI deployments for health data processing.
PepsiCo
Clustering and recommendation model with TiB-scale data pipeline. Auto-switching between Pandas and Spark based on data volume using Databricks and Snowflake.
Rhythmic Science
HIPAA-compliant data de-identification pipeline with BERT-based NLP for health report processing. Dockerized and deployed on AWS EC2 with FastAPI.
Pfizer
Optimized and built numerous ETL pipelines consuming data from Snowflake with PostgreSQL ingestion. Reduced pipeline cost by 2x through transaction optimization.
PlusPower
Comprehensive SageMaker pipeline for data processing, XGBoost training, evaluation, and deployment. Increased test coverage from 20% to 85% with integration tests.
NewsCorp
ChatGPT-like experience with open-source LLMs (LLaMA, Falcon, Stable Diffusion). Models deployed on EC2 with FastAPI and TGI for optimized latency and throughput.
Siemens Energy
Multi-layer ETL pipeline with 15+ Spark jobs on AWS Glue orchestrated by Airflow. Landing, staging, enrichment, and final layers processing data across S3, Aurora, and Snowflake.
Lifebit
End-to-end automated MLOps workflow: dockerization, deployment, evaluation, health checks, and zero-downtime model promotion. Reduced cost 3x and increased throughput from 30K to 200K articles/day.
SmartCat
Real-time computer vision system with YOLO object detection and custom Keras classifiers. End-to-end MLOps with MLFlow, LakeFS, and TensorFlow Serving on AWS.
SmartCat
Analytics pipeline calculating business KPIs from computer vision outputs using Spark (Scala) for distributed processing, orchestrated with Airflow and Prefect.
HTEC
COVID-19 test detection and classification model trained and deployed on mobile devices for real-time inference.
AMD
R&D for game super-resolution model (similar to NVIDIA DLSS). Custom loss functions, layers, and metrics implemented from scratch in PyTorch from research papers.
Trusted By
Let's Talk
Have a project in mind? We'd love to hear about it.