Work Experience
Senior AIOps Manager - Contract - Tredence (Copilot (Agentic Contract/NDA Assistant on Databricks + Copilot Studio))
Duration: Nov 2025 – Present
Summary:- Developed an enterprise-grade multi-agent AI assistant for contract and NDA analysis using LangGraph and Azure OpenAI services, deployed on Databricks
- The system enables natural language clause lookup, summarization, and metadata-aware document retrieval to streamline contract management workflows
Responsibilities:- Engineered multi-agent LangGraph pipeline for intent detection, classification, semantic retrieval, and response generation.
- Integrated Azure OpenAI, Azure AI Search, and SharePoint for enhanced document search and retrieval.
- Deployed scalable MLflow PyFunc model with versioning and serving endpoints.
- Implemented metadata extraction with Azure Function Apps and OCR.
- Enabled conversational contract analysis within Microsoft Teams and Power Platform workflows.
Technologies: LangGraph, Azure OpenAI (GPT-4.1 + embeddings), Azure AI Search (HNSW + OCR + hybrid semantic search), SharePoint, Databricks MLflow PyFunc, Python, Azure Function Apps, Microsoft Copilot Studio, REST APIs
Senior AIOps Manager - Contract - Tredence (Sales Agent NL2SQL System (Agentic AI Assistant for Sales Performance Management))
Duration: Nov 2025 – Present
Summary: Designed and implemented an agentic natural language to SQL analytics system to enable non-technical sales stakeholders to query sales performance data using natural language, improving accessibility and reducing BI turnaround times.
Responsibilities:- Developed multi-stage LangGraph orchestration pipeline for query expansion, intent extraction, schema mapping, SQL generation, execution, and natural language response.
- Integrated Azure OpenAI for intent classification and SQL synthesis, and Databricks Vector Search for schema retrieval.
- Built robust SQL generation pipeline ensuring schema-correct Spark SQL output.
- Packaged workflow as MLflow pyfunc model with secure inference.
- Generated business-readable narrative summaries from query results.
Technologies: LangGraph, Azure OpenAI (GPT-4), Databricks Vector Search, Spark SQL, MLflow pyfunc, Python, Azure AD/MSAL
Senior AIOps Manager - Contract - Tredence (AI-Powered Real-Time Tutoring Platform (GROW Classroom – Paper Education))
Duration: Nov 2025 – Present
Summary: Architected a full-stack AI tutoring system enabling real-time voice-based lessons with adaptive instruction and multi-agent AI orchestration, supporting interactive online learning experiences.
Responsibilities:- Developed AI agent orchestration with LLM providers, STT, and TTS using a provider factory architecture.
- Built low-latency conversational pipelines integrating VAD, turn detection, streaming STT/TTS, and real-time communication.
- Implemented session state orchestration, persistence services, and AI safety monitoring infrastructure.
- Delivered scalable full-stack web interface with Next.js, TypeScript, React, and LiveKit SDK.
Technologies: Python, LiveKit WebRTC, Next.js, TypeScript, React, Google Gemini, OpenAI, Deepgram STT, ElevenLabs TTS, Silero VAD, Pydantic, Google Cloud Storage, Slack
Senior AIOps Manager - Contract - Tredence (MCP File Edit Server for Claude Desktop)
Duration: Nov 2025 – Present
Summary: Built a FastMCP-powered file server to enable Claude Desktop AI to manipulate remote and local files asynchronously, supporting code analysis, editing, and repository operations over SSH.
Responsibilities:- Implemented abstraction layer for local and SSH file systems with AST parsing, regex, and Git integration.
- Registered over 30 MCP tool operations with error handling and safety validation.
Technologies: FastMCP, Python, AST parsing, Git, SSH
Senior AIOps Manager - Contract - Tredence (Low-Resource ASR/NMT (Tamasheq – Digital Prybar))
Duration: Nov 2025 – Present
Summary: Developed an offline-capable speech-to-text and translation pipeline for low-resource languages, integrating diarization and noise reduction to support offline portability and human-in-the-loop evaluation.
Responsibilities:- Built ASR and NMT pipeline using Facebook MMS-1B and NLLB-200 models.
- Implemented diarization and noise reduction components.
- Delivered API and GUI interfaces containerized with Docker.
- Integrated human-in-the-loop evaluation workflows.
Technologies: Facebook MMS-1B, NLLB-200, pyannote.audio, Demucs, Silero VAD, Flask REST API, PyQt, Docker, Label Studio
Senior AIOps Manager - Contract - Tredence (Flexible GraphRAG Platform (Hybrid RAG + Knowledge Graph))
Duration: Nov 2025 – Present
Summary: Engineered a modular hybrid retrieval-augmented generation (RAG) platform combining document processing, knowledge graph construction, and multi-database hybrid search with LLMs for enterprise deployments.
Responsibilities:- Designed FastAPI microservices for asynchronous ingestion and real-time progress updates.
- Supported multiple data formats and integrated vector and graph databases.
- Delivered three full frontends using React, Angular, and Vue.
Technologies: FastAPI, OpenAI, Google Gemini, LlamaIndex, Qdrant, Weaviate, pgvector, Neo4j, Kuzu, React, Angular, Vue
Senior Machine Learning Engineer - Waste Management (Automated Invoice Processor (Amazon Bedrock + Claude 3.5))
Duration: Jul 2025 – Oct 2025
Summary:- Designed a scalable AI pipeline to extract structured data from compressed natural gas billing PDFs using Claude 3
- 5 Sonnet via Amazon Bedrock, automating billing reconciliation processes
Responsibilities:- Converted PDFs to high-resolution images and parsed content with custom JSON prompts.
- Automated billing reconciliation using AWS S3, Boto3, Pandas, and OpenPyXL.
Technologies: Claude 3.5 Sonnet, Amazon Bedrock, AWS S3, Boto3, Pandas, OpenPyXL, Python
Senior Machine Learning Engineer - Waste Management (Truck Detection and Fraud Analysis)
Duration: Jul 2025 – Oct 2025
Summary: Developed a video analytics pipeline to detect trucks entering and exiting landfill sites and identify fraudulent transactions by matching detections with ticket data.
Responsibilities:- Implemented YOLOv8 and ByteTrack for truck detection and tracking.
- Applied geometric lane mapping, direction validation, and fraud heuristics.
- Integrated with Snowflake and FastLane APIs and deployed multi-camera processing workflows.
Technologies: YOLOv8, ByteTrack, Snowflake, FastLane APIs, Python
Senior Machine Learning Engineer - Waste Management (Integrated Insurance Agentic AI System)
Duration: Jul 2025 – Oct 2025
Summary: Developed a LangChain-based multi-agent AI system automating insurance workflows including document ingestion, validation, communication, and audit tracking to improve operational scalability.
Responsibilities:- Designed agents for OCR, signature check, policy and VIN matching, and customer notifications.
- Implemented secure S3 uploads, Redis/Celery queueing, PostgreSQL JSON models, and tokenized document access with audit logging.
Technologies: LangChain, Django, React, OCR, Redis, Celery, PostgreSQL, AWS S3
Senior Machine Learning Engineer - Waste Management (Flight Booking with Semantic Kernel & AutoGen)
Duration: Jul 2025 – Oct 2025
Summary: Built a multi-agent flight booking system using Microsoft AutoGen and Semantic Kernel, enabling fully automated bookings from natural language instructions.
Responsibilities:- Orchestrated reasoning and action agents for flight search and booking.
- Integrated Azure CosmosDB for real-time flight data storage.
- Developed a web app with Python, Flask, and HTML/CSS.
Technologies: Microsoft AutoGen, Semantic Kernel, Azure CosmosDB, Azure OpenAI (GPT-4), Python, Flask, HTML/CSS
Senior Machine Learning Engineer - Waste Management (Agentic AI Implementation for Financial Crime Compliance (FCC) Automation)
Duration: Jul 2025 – Oct 2025
Summary: Designed and implemented autonomous AI agents to perform adverse media monitoring, sanctions screening, and potential sanctions detection within enterprise compliance workflows for financial crime compliance.
Responsibilities:- Built multi-step AI pipelines using Python, LangChain, spaCy, TensorFlow, PyTorch, and Scikit-learn.
- Developed autonomous data-gathering agents integrating external information providers and web sources.
- Implemented context-aware decision-making logic and explainable AI mechanisms.
- Deployed AI agents on distributed enterprise platform using Java (Spring), RabbitMQ, ELK, Keycloak, and GCP.
Technologies: Python, LangChain, spaCy, TensorFlow, PyTorch, Scikit-learn, Java (Spring), RabbitMQ, ELK, Keycloak, GCP
Senior Machine Learning Engineer - Mashgin (AI/Computer Vision for Touchless Checkout)
Duration: Jun 2021 – Jun 2025
Summary: Developed and optimized computer vision models for Mashgin’s touchless self-checkout kiosks, achieving high accuracy and low latency for retail environments.
Responsibilities:- Developed PyTorch models (e.g., EfficientNet-B4) for item recognition across 60,000+ SKUs.
- Deployed models on NVIDIA Jetson Nano using ONNX Runtime to reduce inference latency.
- Created synthetic data pipeline with Blender and Python for training data generation.
- Implemented 3D reconstruction with Open3D for product digitization.
Technologies: PyTorch, EfficientNet-B4, NVIDIA Jetson Nano, ONNX Runtime, Blender, Python, Open3D
Senior Machine Learning Engineer - Mashgin (MLOps & Cloud Infrastructure)
Duration: Jun 2021 – Jun 2025
Summary: Built and maintained MLOps pipelines and cloud infrastructure to support continuous model updates and reliable deployment across hundreds of kiosks.
Responsibilities:- Developed CI/CD pipelines with GitHub Actions and Kubernetes.
- Designed Redis-based feature store for real-time feature serving.
- Managed AWS infrastructure using Terraform.
- Implemented monitoring with Prometheus and Grafana.
Technologies: GitHub Actions, Kubernetes, Redis, AWS (EC2, S3, SageMaker), Terraform, Prometheus, Grafana
Senior Machine Learning Engineer - Mashgin (Data Engineering)
Duration: Jun 2021 – Jun 2025
Summary: Processed large-scale transaction data and built real-time inventory management APIs to support GDPR-compliant retail deployments.
Responsibilities:- Processed 8TB/day of transaction data using PySpark on Databricks with Delta Lake.
- Streamed transaction events via Apache Kafka.
- Built Node.js REST APIs for real-time inventory management integrated with checkout kiosks.
Technologies: PySpark, Databricks, Delta Lake, Apache Kafka, Node.js, REST APIs
Senior Machine Learning Engineer - Mashgin (Financial Systems)
Duration: Jun 2021 – Jun 2025
Summary: Architected and tested a high-performance payment routing system integrated with contactless payment methods for healthcare and retail environments.
Responsibilities:- Developed Go-based payment router handling over 2 million transactions per day with low latency.
- Conducted performance testing with Locust.
Technologies: Go, Locust
Senior Machine Learning Engineer - Mashgin (Sales Analytics & Forecasting (Time Series ML Project))
Duration: Jun 2021 – Jun 2025
Summary: Led development of a sales forecasting system for over 1,100 retail partner locations using advanced time series machine learning techniques.
Responsibilities:- Implemented DTW-based time series clustering and engineered temporal, promotional, holiday, and store-specific features.
- Built stacking ensemble models using XGBoost, Gradient Boosting, Linear Regression, Random Forests, and Decision Trees.
- Created full ML pipeline for feature engineering, model training, evaluation, and batch inference.
Technologies: Python, scikit-learn, XGBoost, tslearn
Senior Machine Learning Engineer - Mashgin (LLM & NLP Development)
Duration: Jun 2021 – Jun 2025
Summary: Fine-tuned large language models and built retrieval-augmented generation pipelines to support research use cases on large academic and internal document collections.
Responsibilities:- Fine-tuned GPT-Neo using Hugging Face Transformers and DeepSpeed for decentralized training.
- Built RAG pipelines with LlamaIndex and Weaviate for efficient querying.
Technologies: GPT-Neo, Hugging Face Transformers, DeepSpeed, LlamaIndex, Weaviate
Senior Machine Learning Engineer - Mashgin (GPU Programming & AI Framework Optimization)
Duration: Jun 2021 – Jun 2025
Summary: Optimized GPU training and inference pipelines to reduce training time and improve inference performance for AI models.
Responsibilities:- Applied CUDA C++ kernels and mixed precision training with cuDNN and AMP.
- Optimized inference with TensorRT and benchmarked distributed multi-GPU training with NCCL.
- Profiled performance using NVIDIA Nsight and deployed inference pipelines with Triton Inference Server.
Technologies: CUDA C++, PyTorch, cuDNN, AMP, TensorRT, NCCL, NVIDIA Nsight, Triton Inference Server
Senior Machine Learning Engineer - Mashgin (Decentralized ML Infrastructure)
Duration: Jun 2021 – Jun 2025
Summary: Contributed to decentralized machine learning infrastructure including task distribution, distributed training, and P2P networking for scalable deep learning workloads.
Responsibilities:- Implemented Rust-based task distribution system using libp2p.
- Developed distributed training framework with PyTorch and Ray.
- Designed Scala pipelines for Spark ML workloads.
- Developed Go-based P2P networking layer for model sharing.
Technologies: Rust, libp2p, PyTorch, Ray, Scala, Spark ML, Go
Machine Learning Engineer | MLOps Lead - Voyage (Autonomous Vehicle Perception)
Duration: Jan 2017 – May 2021
Summary: Developed perception models and pipelines for autonomous shuttles operating in retirement communities, improving detection accuracy and real-time navigation.
Responsibilities:- Trained YOLOv3 models on Azure ML for pedestrian and obstacle detection.
- Optimized LiDAR point cloud processing with PointNet and CUDA.
- Developed OpenCV-based object detection pipelines.
Technologies: YOLOv3, TensorFlow, Azure ML, PointNet, CUDA, OpenCV
Machine Learning Engineer | MLOps Lead - Voyage (MLOps & Data Engineering)
Duration: Jan 2017 – May 2021
Summary: Built data pipelines and orchestration systems to support large-scale sensor data processing and automated model retraining for autonomous vehicle fleet.
Responsibilities:- Built Spark pipelines on Databricks processing 25TB/month of sensor data.
- Orchestrated Airflow DAGs with Kubernetes for automated retraining.
- Implemented MLflow for experiment tracking and model versioning.
Technologies: Spark, Databricks, Airflow, Kubernetes, MLflow
Machine Learning Engineer | MLOps Lead - Voyage (Edge Deployment & Simulation)
Duration: Jan 2017 – May 2021
Summary: Deployed models to edge devices and extended simulation environments to improve autonomous shuttle training and perception.
Responsibilities:- Deployed TensorFlow Lite models achieving low inference times on edge devices.
- Extended CARLA simulator with Unreal Engine plugins.
- Developed C++ modules for sensor fusion integrating multiple sensor data types.
Technologies: TensorFlow Lite, CARLA, Unreal Engine, C++
Machine Learning Engineer | Full-Stack Developer - Startup X (Augmentus (Robotics Programming Platform))
Duration: Jan 2015 – Dec 2016
Summary: Developed robotics programming platform enabling no-code robotic automation for industrial applications with real-time computer vision and visualization.
Responsibilities:- Developed ROS nodes in Python and C++ for robotic path planning.
- Integrated OpenCV for real-time object detection.
- Built WebGL visualization dashboard using Three.js.
Technologies: ROS, Python, C++, OpenCV, WebGL, Three.js
Machine Learning Engineer | Full-Stack Developer - Startup X (XpertFlow (Healthcare AI))
Duration: Jan 2015 – Dec 2016
Summary: Deployed machine learning models for ECG signal analysis and designed HIPAA-compliant data storage supporting secure processing of large patient datasets.
Responsibilities:- Deployed TensorFlow models for arrhythmia detection.
- Designed PostgreSQL schemas for HIPAA-compliant patient data storage.
- Implemented Python preprocessing pipelines with NumPy and SciPy.
Technologies: TensorFlow, PostgreSQL, Python, NumPy, SciPy
Machine Learning Engineer | Full-Stack Developer - Startup X (StaffAny (Workforce Management SaaS))
Duration: Jan 2015 – Dec 2016
Summary: Scaled backend and developed frontend for workforce management platform handling time-tracking and scheduling for thousands of users.
Responsibilities:- Scaled Node.js backend with Express and MongoDB.
- Developed React frontend for management dashboards.
- Implemented RabbitMQ for asynchronous task processing.
Technologies: Node.js, Express, MongoDB, React, RabbitMQ
Research Intern – Machine Learning - Amazon (Product Recommendation Systems)
Duration: Jun 2014 - Dec 2014
Summary: Developed machine learning models and ETL pipelines to improve product recommendation accuracy on Amazon’s e-commerce platform.
Responsibilities:- Developed recommendation models using scikit-learn and Mahout.
- Analyzed large-scale customer behavior data.
- Built ETL pipelines with AWS Data Pipeline.
Technologies: scikit-learn, Mahout, Pandas, Matplotlib, AWS Data Pipeline
Education
- University of Central Florida
Master of Science in Computer Science
May 2012 – May 2014
- University of Central Florida
Bachelor of Science in Computer Science
Sep 2007 – Jun 2011
Upstaffer since March, 2026
