Agentic AI Systems
Designing task-driven AI agents with memory, tool use, orchestration, and human-in-the-loop controls.
Freelance AI Engineer
Ready to ship intelligent AI systems that work.
I design and deploy AI-first products, automation engines, and ML pipelines for teams that want speed, privacy, and measurable impact.
- 15+ Production-grade AI workflows
What I Do
Freelance Services
ML Product Engineering
End-to-end model development, evaluation, and deployment with robust APIs and production monitoring.
Data + Automation Pipelines
Airflow, dbt, cloud-native ETL, and analytics workflows for clean, scalable decision infrastructure.
Computer Vision Applications
Real-time edge-friendly systems for pose tracking, gesture intelligence, and automation-ready visual inputs.
Proof of Delivery
Previous Client Work
Mavonics
Engineered a real-time wrist pose-tracking pipeline for a handwriting product using TensorFlow Lite and React Native.
Nova Retail Labs
Built a demand forecasting stack and automated reporting agents to reduce weekly ops analysis effort.
Water Facility Management Client
Designed a classical ML time-series anomaly detection system for equipment monitoring (pumps, level switches), with full data privacy and on-premise deployment.
Foundry Health Data
Built a multi-agent AI clinic assistant with appointment scheduling, patient triage, and admin tools—deployed on AWS. Reduced manual admin work by 45+ minutes daily.
Selected Build Log
Projects and Portfolio
Computer Vision
Focus Play: AI-Powered Productivity Tool
Built a local-first OpenCV + MediaPipe system that pauses and resumes media playback based on user presence.
- Low-latency inference pipeline optimized for varying lighting conditions.
- Privacy-first architecture with minimal CPU footprint.
ML Infrastructure
AutoImpute: Smart Data Imputation Library
Created a Python package for intelligent missing-value treatment across mixed data types.
- Parallel processing and caching for large-scale datasets.
- CI/CD with GitHub Actions and PyPI distribution automation.
Data Engineering
Subreddit Analysis Platform
Deployed an ETL stack processing high-volume Reddit data with Airflow, AWS Redshift, and dbt.
- Containerized workflows and Infrastructure as Code for low-latency pipelines.
- Real-time sentiment insights delivered through Streamlit dashboards.
Predictive Modeling
Bank CD Subscription Predictor
Developed an interpretable XGBoost model to estimate subscription likelihood in banking campaigns.
- Balanced class imbalance with SMOTE and weighted loss functions.
- SHAP-based feature attribution for explainable business decisions.
Anomaly Detection
Water Management: Flow & Consumption Prediction
Built a classical ML time-series anomaly detection system for water facility management, tracking pump cycles and level switches.
- Signal-to-behavior feature engineering: pump runtime durations, idle patterns, daily/weekly cycle analysis.
- Privacy-first architecture deployed on client cloud (AWS/Azure), no external API calls or AI model training on sensitive data.
- Anomaly detection flags low-flow periods and water unavailability patterns automatically.
AI Agents for Healthcare
Pulmonary Care AI Assistant: Multi-Agent Clinic System
Engineered an intelligent triage and scheduling agent for small clinics using LangGraph + LangChain orchestration with Streamlit frontend.
- Agent Nodes: Symptom triage (LLM-driven), appointment booking (calendar sync), patient records (memory store), admin alerts (notifications).
- Tech Stack: LangGraph for workflow orchestration, OpenAI/Groq for LLM inference, FastAPI backend, Streamlit UI, deployed on AWS free tier.
- Core Features: Appointment scheduling with SMS/WhatsApp reminders, patient profiles with history/allergies, emergency routing, billing summaries, inventory checks, no-show reduction.
How I Work
Delivery Process
- Scope Sprint: Define business objective, success metrics, and data constraints.
- System Blueprint: Architect agent flow, model stack, and deployment path.
- Rapid Build: Ship an iterative MVP with measurable milestones.
- Production Hardening: Observability, optimization, and long-term maintainability.
Open for Freelance
Let us build your next AI product.
Available for contract and project-based engagements in agentic AI, machine learning systems, and data platform development.