I am an AI/ML and Senior Software Engineer with 5+ years of experience building cloud-native applications, distributed data systems, and applied AI products across healthcare, media streaming, and consulting.

My recent work focuses on:

• Generative AI and RAG: grounded search, hybrid retrieval, reranking, citations, hallucination controls, and evaluation.
• Document intelligence: OCR, metadata extraction, PHI/PII processing, multimodal enrichment, and semantic indexing.
• AI platform engineering: asynchronous APIs, event-driven workers, model-serving workflows, observability, and secure cloud deployment.
• Full-stack systems: Python/FastAPI services, React/Next.js interfaces, PostgreSQL, MongoDB, Kafka, Redis, and cloud infrastructure.

I enjoy turning ambiguous product requirements into reliable systems with clear APIs, measurable performance, and production-focused engineering practices.

• LLM applications, RAG, agent workflows, prompt orchestration, and structured outputs
• BM25 and dense retrieval, FAISS, pgvector, Milvus, Elasticsearch/OpenSearch
• Transformers, PEFT/LoRA, embeddings, reranking, evaluation, and grounding
• OCR, NLP, entity extraction, document classification, OpenCV, Tesseract, spaCy, and Whisper

• FastAPI, Azure Functions, REST/GraphQL APIs, asynchronous workers, Celery, and Pub/Sub
• PostgreSQL, MongoDB, DynamoDB, Redis, Kafka, Spark, and data pipelines
• Azure Blob Storage, Service Bus, Azure ML, GCS, Cloud Run, Vertex AI, Lambda, S3, and ECS

• Docker, Kubernetes, Helm, Terraform, GitHub Actions, and CI/CD
• OpenTelemetry, Prometheus, Grafana, structured logging, tracing, and correlation IDs
• Pytest, Hypothesis, React Testing Library, Vitest, and Playwright

August 2023 – October 2025

• Built and scaled an intelligent document-processing platform handling approximately 1 million PHI/PII documents across PDFs and images.
• Reduced OCR and extraction latency from roughly 2 minutes to about 15 seconds through asynchronous processing, optimized computer-vision workflows, and cloud-native orchestration.
• Designed non-blocking ingestion APIs using Azure Functions, Blob Storage, Service Bus, managed identity, idempotency controls, and correlation IDs.
• Developed document enrichment and retrieval workflows using OCR, NLP, transformers, embeddings, vector search, and metadata extraction.
• Built reusable RAG and vector-search components using technologies such as Milvus, PostgreSQL/pgvector, Elasticsearch/OpenSearch, and MongoDB.
• Improved platform reliability through structured logging, distributed tracing, metrics, automated testing, and secure PHI/PII handling.
• Contributed to containerized deployment, Kubernetes workflows, CI/CD automation, and ML lifecycle tooling.

October 2021 – June 2023

• Contributed to backend and data-platform services supporting HBO Max content distribution and metadata workflows.
• Helped migrate more than 1 million content identifiers into DynamoDB-backed services using AWS Lambda and Kubernetes-based workloads.
• Built integrations supporting distribution across Google, Apple, Roku, and Amazon ecosystems and international market expansion.
• Developed event-driven metadata-processing workflows using AWS services, Kafka, Spark, S3, and asynchronous service communication.
• Improved operational visibility with Grafana dashboards, centralized logging, alerting, and incident-response integrations.
• Optimized large metadata feeds and parallel processing paths to improve throughput and reduce processing time.

March 2021 – October 2021

• Built full-stack cloud applications and automation tools using AWS Lambda, DynamoDB, Step Functions, Node.js, Angular, and Python.
• Contributed to the me@slalom internal platform, improving project-assignment and employee workflow efficiency.
• Developed REST APIs, Elasticsearch integrations, data pipelines, and Redshift-oriented processing workflows.
• Created automated testing and delivery workflows using Selenium and GitHub Actions.

May 2017 – July 2018

• Supported database modernization and reporting initiatives across student-information and operational systems.
• Migrated and consolidated data from multiple legacy systems into Oracle-based platforms.

A pre-alpha, Apache-2.0 toolkit connecting computer-vision road observations, traffic forecasting, privacy-aware GPS processing, GeoJSON, map-matching contracts, and route-reliability scoring.

Recent work: native Go execution with Python conformance fixtures, portable linear-model inference trained in Python, experimental neural traffic forecasts with reproducible baselines, and per-class detection metrics. The project provides reusable components; a trained detector and hosted routing service remain roadmap items.

Highlights: Python, Go, scikit-learn, time-series backtesting, GPX/GeoJSON, consent-gated trace processing, and cross-language testing.

A local-first, full-stack platform that processes video into metadata, sampled frames, scene boundaries, transcripts, structured AI tags, embeddings, human review, and semantic discovery using open-source models.

Highlights: Next.js 15, React 19, FastAPI, PostgreSQL/pgvector, Celery/Valkey, FFmpeg, Ollama, Qwen2, Granite Vision, faster-whisper, and Docker Compose. GCP deployment and cloud adapters are optional.

A multi-agent application that converts natural-language API requirements into persisted plans and generated artifacts through a planner → coder → tester → reviewer → deployer workflow.

Highlights: FastAPI, Next.js, PostgreSQL, Redis, Celery, JWT authentication, Alembic, structured agent-run persistence, Prometheus/Grafana, Jaeger, OpenTelemetry, Playwright, and CI/CD.

A modular retrieval-augmented generation system with document ingestion, BM25 and FAISS hybrid retrieval, reranking, source confidence, answerability checks, citation validation, claim verification, versioned APIs, and evaluation hooks.

Highlights: FastAPI, SentenceTransformers, FAISS, BM25, OpenAI-compatible generation, tenant-aware APIs, hallucination controls, and golden-set evaluation workflows.

A privacy-focused browser application that imports Snapchat export media directly into Google Photos without sending user files through a hosted backend.

Highlights: React, Vite, Google Identity Services, Google Photos Library API, browser-side ZIP processing, resumable-upload recovery, rate-limit backoff, duplicate tracking, and downloadable JSON/CSV reports.

A Flask dashboard for public seismic observations, source-aware caching, interactive maps, a 3D globe, and experimental probabilistic risk indicators.

Recent work: coordinate normalization across data providers, source/feed filtering, separation of demo records from real observations, and antimeridian-safe map queries.

Highlights: Python, Flask, SQLite, Chart.js, Leaflet, globe.gl, USGS/GeoJSON, and scikit-learn. Forecasts are educational indicators, not exact earthquake predictions.

A modular analytics API and Streamlit dashboard for seismic filtering, geospatial queries, anomaly detection, and spatial clustering.

Recent work: spatial hash-grid neighbor lookup for DBSCAN, query and memory-use improvements, and clearer dashboard diagnostics.

Highlights: Flask, SQLite, Pandas, Streamlit, Plotly, Docker, and GitHub Actions.

A Python CLI and Flask application for extracting filesystem and embedded metadata from PDF, DOCX, XLSX, PPTX, and legacy Office documents.

Highlights: reusable extraction services, safe JSON serialization, temporary-file cleanup, graceful dependency handling, and Render deployment support.

An end-to-end transformer fine-tuning and inference project that trains GPT-2 on domain-specific text and serves generation through FastAPI with a lightweight frontend.

Highlights: Hugging Face Trainer, PyTorch, FastAPI, model fallback behavior, health monitoring, and deployment configuration.

I build and maintain my own public projects, contribute features and reliability fixes, and integrate community contributions. Selected merged work:

Interested in collaborating? Explore the OpenTrace ML contribution guide and open issues.

• PHI/PII Document Extraction Service: asynchronous Azure architecture for secure uploads, queue-based processing, OCR, NLP enrichment, idempotency, and high-volume extraction.
• Vertex AI Search and Chat Services: shared RAG pipeline using Gemini, grounded search, structured responses, and cloud-native service APIs.
• Enterprise Semantic Search: hybrid keyword and vector retrieval across enriched document metadata with reusable search abstractions.
• ML Platform and Serving: containerized inference services, model evaluation, telemetry, CI/CD, and scalable cloud deployment patterns.
• Media Metadata Platforms: event-driven content processing, partner integrations, catalog migration, and operational dashboards for streaming systems.

Some enterprise projects are proprietary; descriptions focus on architecture, responsibilities, and measurable outcomes without exposing confidential implementation details.

• M.S. in Computer Science, Oklahoma Christian University — GPA: 4.0
• M.S. in Computer Science, The University of Texas at Arlington
• B.E. in Information Technology, Vadodara Institute of Engineering

My recent projects reflect the areas I want to keep exploring:

• Applied ML, computer vision, and geospatial intelligence: road observations, map matching, GPS/GeoJSON, and route scoring through OpenTrace ML.
• Time-series forecasting and ML evaluation: incremental learning, neural baselines, rolling backtests, and reproducible comparisons.
• Generative AI, RAG, and agent workflows: grounded retrieval, reranking, citations, hallucination controls, and multi-agent API generation.
• Multimodal and local AI: video understanding, speech-to-text, structured tagging, embeddings, and human review through CineTag.
• Distributed systems and platform engineering: asynchronous APIs, queues, observability, reliable processing, and cloud deployment.
• Data engineering and interactive analytics: source normalization, spatial query optimization, clustering, and map-based visualization.
• Open-source engineering: reusable Python/Go components, contributor collaboration, documentation, testing, and CI.

• Status: actively searching for a new role.
• Roles of interest: AI/ML Engineer, AI Software Engineer, and Software Engineer roles focused on AI applications, backend services, or cloud platforms.
• Time-zone flexibility: available to work across any time zone.
• Relevant experience: intelligent document processing at CVS Health, streaming metadata and distributed services at Warner Bros. Discovery / HBO Max, and full-stack cloud applications at Slalom Consulting.
• Contact: LinkedIn or email.

• LinkedIn: linkedin.com/in/vrajpatell0712
• Portfolio: vrajpatell.github.io/portfolio
• Email: vrajpatell0712@gmail.com
• GitHub: github.com/vrajpatell