Hire the Best Facial Recognition Specialists

Top 1% Expert-Vetted Freelancer. I have a PhD in Computer Science (AI & Machine Learning) and I am a physician. I build custom AI systems with proprietary and public data, including creation of novel datasets and state-of-the-art models. * After a 30-minute to 1-hour initial meeting, I'll craft a clearly written report outlining a unique AI strategy for your project. * For long-term engagements, I implement AI solutions end-to-end, including data collection, data cleaning, novel model design, model implementation and training in Python, and iteration on the modeling and data strategy to achieve excellent performance. * One of my specialties is creating custom models for proprietary datasets, including images, videos, audio recordings, and domain-specific data. Book a 30-minute consultation with me to get immediate answers about your AI project: - How can you leverage AI in your organization? - What kind of AI do you need to solve a particular problem? - Should you use a model like ChatGPT/Claude or do you need a custom model? - What modeling strategy should you use? - How should you collect and annotate data? - How much data will you need? - How should you measure performance? After a consultation I’ll craft a clearly-written report outlining an AI strategy for your project. Specialties: Healthcare AI: I am an expert in healthcare AI, with a unique background that blends clinical understanding from my MD, deep technical expertise from my AI PhD, and business experience. I have completed 100+ engagements across medical specialties including radiology, plastic surgery, orthopedic surgery, dermatology, cardiology, mental health, family medicine, dentistry, physical therapy, women’s health, pediatrics, nutrition, and neurology. I have created datasets and custom models for diverse forms of biomedical data including x-rays, CTs, MRIs, clinical photographs, patient videos, medical audio, medical notes, EHR data, insurance claims, omics data, and more. Computer Vision: I create custom AI models for 2D images, 3D images, and videos on proprietary datasets. I am excited about projects from multiple industries including healthcare, manufacturing, automotive, and agriculture. I have extensive experience with classification, object detection, segmentation, and keypoint detection, to identify and localize abnormalities or features of interest. I developed the first machine learning model in the world to predict multiple abnormalities simultaneously from a CT scan. For a dermatology practice, I built the first CV model to predict Fitzpatrick skin type, pigmentation, redness, and wrinkle severity (mean accuracy 85%)—now a core model at Kesty AI. Artificial Intelligence R&D: I have led multiple AI research projects across industry and academia. I have developed novel AI methods, including HiResCAM, a convolutional neural network explanation method with mathematical guarantees. I have published original research across multiple areas of AI, including computer vision, natural language processing, explainable AI, expert systems, and applied AI. My research papers have been cited over 1,000 times. My healthcare AI blog Glass Box has over 700,000 readers from 185 countries. Natural Language Processing (NLP): I have a deep understanding of large language models (LLMs) like Claude, Gemini, ChatGPT, and Llama. I have leveraged Transformers and other NLP techniques for numerous applications, including customized chatbots, medical note generation, and structured information extraction. Advising Entrepreneurs: Before focusing on AI consulting, I spent seven years as the founder of a health AI startup. I led my previous company from concept to deployed B2B SaaS product serving medical practices. Our AI history-taking assistant and AI scribe saved clinicians 2+ hours daily. I managed engineering teams of 5-10 (60+ contributors), secured two U.S. patents, and raised competitive grant funding. I enjoy working with entrepreneurs and discussing pitch decks, fundraising, customer discovery, designing an MVP, and evaluating the ROI of an AI product. If you’d like to talk with me about your AI project, please feel free to send me a message or book a consultation using the link on my profile.

Do you have an AI vision that needs to become a real, working product? I don't just build models; I engineer complete, scalable solutions that turn data into actionable insights and automation. For over five years, I've specialized in bridging the gap between cutting-edge Artificial Intelligence (AI) research and robust software that delivers real-world value. My core expertise lies in computer vision and machine learning, but my skill set is full-stack. This means I can own your project from the initial data pipeline, through model training and optimization, all the way to deploying a polished desktop application or a secure enterprise API. I thrive on building tools that work seamlessly for end-users, whether it's a retail manager, a traffic controller, or a sports coach. My strongest suit is developing intelligent systems that "see" and understand the world. I've built a retail analytics platform (CrowdIQ) that transforms standard CCTV into a source of business intelligence, tracking customer demographics and behavior. In the sports domain, I created PadelIQ, an analytics engine that uses computer vision to track player movement, posture, and court coverage from match footage, providing real-time coaching feedback. For public safety, I developed a traffic management system (OmniRoad AI) using advanced object detection for real-time accident and congestion monitoring. Beyond computer vision, I architect full-scale data science pipelines. A prime example is my telecom churn prediction project, where I built a machine learning model to identify at-risk customers and paired it with an interactive Power BI dashboard. This end-to-end approach—from data analysis to a clear visualization of insights—ensures the model's findings directly inform business strategy and retention actions. I also develop the tools and infrastructure that power AI applications. I've built secure, enterprise-grade systems like DevelmoGPT, a RAG-based LLM that allows for secure, semantic search over private company documents. From creating simple utilities like PDF-to-audio converters to designing complex role-based access systems, I ensure the foundation of any AI solution is reliable, secure, and maintainable. My process is collaborative and results-driven. I start by deeply understanding your business problem, not just the technical requirement. We'll then iterate through prototyping, development, and testing to ensure the final product not only meets specs but also delivers tangible ROI. I communicate clearly at every stage, providing demos and documentation so you're never in the dark. Let's connect. Share your project idea or challenge, and I'll provide a clear outline of how we can leverage AI, machine learning, or computer vision to build your intelligent solution. Click the invite button to start the conversation. /// The following is just for SEO. You can ignore it /// #computer vision #computer vision engineer #computer vision OpenCV #machine learning computer vision #deep learning computer vision #computer vision machine learning #machine learning python #nlp machine learning

I design, optimize, and integrate real-time object detection and tracking pipelines for NVIDIA Jetson, RK3588, and cloud environments — with measurable gains in FPS, latency, and deployment stability. From YOLO training to TensorRT optimization and production integration, I build Computer Vision systems that work reliably on real hardware. If you need more than just a trained model — if you need a working AI system integrated into hardware or software — I deliver complete, production-ready solutions. 🎯 WHAT I DELIVER • End-to-End Deep Learning Pipelines Model architecture → dataset optimization → training → evaluation → deployment • Real-Time Object Detection & Multi-Object Tracking Optimized YOLO pipelines with stable tracking (DeepSORT / ByteTrack / BOT-SORT) • ⚡ Edge AI Acceleration & Performance Optimization TensorRT conversion, CUDA acceleration, latency reduction, memory tuning, FPS improvements • 🔗 AI Integration into Production Systems Jetson deployment, inference APIs, embedded integration, debugging, monitoring & system optimization 🏆 PROVEN EXPERIENCE • 5★ Upwork reviews for NVIDIA Jetson Nano, AGX & Orin deployments • Deployed Frigate and real-time CV pipelines on Jetson Orin NX • Optimized inference performance using TensorRT for improved real-time execution • Installed and configured LLM environments with secure key management & usage tracking • Research & production AI deployment experience at HBrain 🛠 TECH STACK Deep Learning: PyTorch, CNN architectures, YOLO variants Computer Vision: OpenCV, real-time video processing, detection & tracking Optimization: TensorRT, CUDA Systems: Linux, Embedded Systems, NVIDIA Jetson Languages: Python, C++ If you share your hardware target, dataset sample, or performance goal, I can propose a clear technical architecture and realistic delivery plan.

Machine Learning and AI Engineer. 8+ years shipping applied AI into production: LLM agents and RAG, NLP, speech, computer vision, and CUDA. I build systems that hold up in production, not demos, and every claim below is a number I measured. LLM and NLP - Semantic CV-to-job matching for an EU job board: 10,001 CVs against 10,003 vacancies, 100M cosine pairs in about 18s on one RTX 4090 (bge-m3), then a two-stage LLM re-rank that spreads the top 10 from 0.93 down to 0.71 where the raw cosine moves only 0.03. - LLM CV builder: messy PDF / DOCX / LinkedIn text into a strict Pydantic record bound to the client's 674-value taxonomy, so invalid output is structurally impossible. Emits a recruiter-ready LaTeX PDF and database-ready JSON. - Crypto signals from tweet embeddings (live client program): Qwen3-Embedding-8B on an A40, 1M+ tweet vectors in Qdrant, 44 symbols scored every 60 seconds, 4.65M events ingested, 10-day zero-error run. Speech - On-prem call-centre speech-to-text, in production 2020-2022, feeding fraud detection and call routing. Benchmarked on hand-labeled calls: Whisper 20% median WER, Vosk 28%, Google 32%. - Diarized transcription toolkit, TTS, and cloned-voice agents. Computer vision and CUDA - Multi-camera 3D capture for film relighting: 33 datasets delivered, 3-32 synchronized cameras each, GPU space carving at 1.98 s/frame versus 65 s in vectorized NumPy (about 33x). 5.0 Upwork review. - GPU brute-force passphrase recovery: about 58,000 candidates/s on one RTX 4090, 160x over CPU, sharded across multiple GPUs and machines. - Open-set retail robot vision on ROS 2; banknote detection at the cash desk. YOLO, Grounding DINO, RealSense. How I work Anti-hallucination is enforced in the code path, not in a prompt: strict schema validation, retry on violation, closed enum vocabularies. Success criteria are pre-registered before results exist, models are judged on time-ordered splits, and negative results get reported rather than buried. Stack - LLM and agents: OpenAI API, Claude Agent SDK + MCP, LangChain / LangGraph, RAG, vLLM, Hugging Face Transformers, fine-tuning, structured outputs with Pydantic + instructor, hand-written tool-use loops, chatbots on Telegram and web - NLP and search: embeddings, semantic search, vector databases (Qdrant), clustering, topic discovery, classification, multilingual pipelines - Speech: Whisper, Vosk / Kaldi, Deepgram, Google Cloud STT, TTS, speaker diarization - Vision: OpenCV, YOLO, Grounding DINO, TensorFlow Object Detection, RealSense RGB-D, ROS 2, NVIDIA Jetson - Classic ML: PyTorch, TensorFlow, scikit-learn, CatBoost, forecasting, time-series validation, reinforcement learning - GPU: CUDA C++, PyCUDA, Nsight Compute, multi-GPU sharding - Data and infra: Python, FastAPI, Docker Compose, PostgreSQL, Redis, BigQuery, pandas / NumPy / Parquet, n8n, Linux VPS, GCP Full case studies for each project above, with architecture diagrams and honest results including what did not work, are in my Portfolio section below. Tell me what you are building and I will tell you plainly whether I am the right fit for it.

Computer Vision Engineer building real-time video analytics, object detection, image segmentation, pose estimation, OCR, and industrial inspection systems. I use OpenCV, YOLO, PyTorch, ONNX Runtime, and TensorRT to take models from data to deployment. I build practical Computer Vision systems for products and operations where accuracy, latency, and reliability matter. ✅ 3+ years of hands-on Computer Vision and Machine Learning experience ✅ 100% Job Success Score and consistent 5.0-star client feedback ✅ Real-time video AI, image processing, detection, segmentation, pose tracking, and OCR ✅ Deployment and inference optimization with ONNX Runtime, TensorRT, CUDA, Docker, FastAPI, Rust, and C++ 𝗖𝗢𝗠𝗣𝗨𝗧𝗘𝗥 𝗩𝗜𝗦𝗜𝗢𝗡 𝗦𝗘𝗥𝗩𝗜𝗖𝗘𝗦 ✅ Object detection and object tracking ✅ Image segmentation, instance segmentation, and mask-based processing ✅ Pose estimation, keypoint detection, posture, movement, and exercise tracking ✅ Real-time camera pipelines and video analytics ✅ OCR, number-plate recognition, and visual data extraction ✅ Industrial visual inspection and defect detection ✅ Image classification, ROI extraction, and dataset preparation ✅ Background removal, image cleanup, and automated image processing ✅ Custom OpenCV, YOLO, PyTorch, and MediaPipe solutions 𝗖𝗩 𝗢𝗣𝗧𝗜𝗠𝗜𝗭𝗔𝗧𝗜𝗢𝗡 & 𝗗𝗘𝗣𝗟𝗢𝗬𝗠𝗘𝗡𝗧 ✅ PyTorch-to-ONNX conversion and ONNX Runtime optimization ✅ TensorRT inference optimization, CUDA acceleration, and latency profiling ✅ NVIDIA Triton Inference Server for scalable model serving ✅ CPU-friendly Computer Vision inference for desktop applications ✅ Rust and C++ support for performance-sensitive CV modules ✅ FastAPI, Docker, Redis, REST APIs, WebSockets, AWS EC2, and Linux deployment 𝗦𝗘𝗟𝗘𝗖𝗧𝗘𝗗 𝗖𝗢𝗠𝗣𝗨𝗧𝗘𝗥 𝗩𝗜𝗦𝗜𝗢𝗡 𝗪𝗢𝗥𝗞 ✅ 𝗥𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝗽𝗼𝘀𝘁𝘂𝗿𝗲 & 𝗺𝗼𝘃𝗲𝗺𝗲𝗻𝘁 𝗮𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 Built a real-time posture-assessment system using custom keypoint-based Computer Vision algorithms. Developed jitter reduction, calibration, head and shoulder pose-state detection, exercise tracking, analytics, and CPU-friendly desktop inference with a C++ inference engine ONNX and MediaPipe. ✅ 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝗶𝗮𝗹 𝘃𝗶𝘀𝘂𝗮𝗹 𝗶𝗻𝘀𝗽𝗲𝗰𝘁𝗶𝗼𝗻 Built an automated automotive fusebox inspection system using 20,000+ images. Delivered ROI extraction, image classification, image segmentation, defect detection, and optimized deployment using PyTorch, ONNX Runtime, TensorRT. Reduced manual inspection from 40 seconds to 1.5 seconds, with optimized processing reduced from 7 seconds to 1.5 seconds. ✅ 𝗠𝘂𝗹𝘁𝗶-𝗰𝗮𝗺𝗲𝗿𝗮 𝘃𝗲𝗵𝗶𝗰𝗹𝗲 𝗮𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 + 𝗢𝗖𝗥 Worked on a multi-camera deployment architecture using optimized TensorRT detection and classification models with OCR on NVIDIA Triton Inference Server. The system identified vehicle type, build, color, and number plates before saving results to a database. ✅ 𝗢𝗡𝗡𝗫 𝗶𝗺𝗮𝗴𝗲 𝘀𝗲𝗴𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 Optimized a high-resolution ONNX background-removal and image-segmentation pipeline forked from Python's rembg package, improving inference performance and reducing image-saving overhead for production use. ✅ 𝗖𝗩-𝗮𝘀𝘀𝗶𝘀𝘁𝗲𝗱 𝗶𝗺𝗮𝗴𝗲 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 Delivered segmentation model debugging and training, 360-degree tripod removal through segmentation-assisted inpainting, trading-card measurement foundations, and custom image-processing workflows. 𝗧𝗘𝗖𝗛𝗡𝗜𝗖𝗔𝗟 𝗦𝗧𝗔𝗖𝗞 Computer Vision: Python • OpenCV • YOLO • PyTorch • TensorFlow • MediaPipe • Object Detection • Object Tracking • Image Segmentation • Image Classification • Pose Estimation • Keypoint Detection • OCR • Video Analytics • Image Processing Optimization: ONNX • ONNX Runtime • TensorRT • NVIDIA Triton Inference Server • CUDA • NVIDIA Jetson • Rust • C++ • CPU/GPU Profiling • Real-Time Inference • Low-Latency Inference Deployment: FastAPI • Docker • Redis • REST APIs • WebSockets • AWS EC2 • Linux • Windows I also have experience with OCR/document AI and other ML workflows, but my primary focus is building, optimizing, and deploying Computer Vision systems. If you need a Computer Vision pipeline built, debugged, optimized, or deployed, send me your current setup and target output. I can help turn it into a reliable system ready for real users.

𝗜 𝗯𝘂𝗶𝗹𝗱 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗿𝗲𝗮𝗱𝘆 𝗔𝗜 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 𝘁𝗵𝗮𝘁 𝗽𝗮𝘆 𝗳𝗼𝗿 𝘁𝗵𝗲𝗺𝘀𝗲𝗹𝘃𝗲𝘀: I saved a fintech client €𝟰𝟬,𝟬𝟬𝟬 𝗽𝗲𝗿 𝘆𝗲𝗮𝗿 by automating their KYC pipeline (𝟵𝟵%+ 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆) and cut clinical documentation time by 𝟲𝟬% with an AI medical scribe inside a live EMR platform. If you have messy documents, manual workflows, or an agent or chatbot idea that needs to actually work in production, not just in a demo, I can ship it fast without sacrificing accuracy or scalability. 𝗪𝗛𝗔𝗧 𝗜 𝗕𝗨𝗜𝗟𝗗 🔹 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀 & 𝗠𝘂𝗹𝘁𝗶 𝗔𝗴𝗲𝗻𝘁 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻: autonomous, decision making agents with LangChain, LangGraph and CrewAI that automate real backend processes: document review, support, lead handling, internal ops. 🔹 𝗥𝗔𝗚 𝗖𝗵𝗮𝘁𝗯𝗼𝘁𝘀 & 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗔𝘀𝘀𝗶𝘀𝘁𝗮𝗻𝘁𝘀: retrieval augmented generation chatbots over your PDFs, docs and databases. Hybrid GraphRAG (Neo4j) plus vector search for grounded, accurate answers with strict guardrails against hallucination. 🔹 𝗢𝗖𝗥 & 𝗗𝗼𝗰𝘂𝗺𝗲𝗻𝘁 𝗔𝗜: invoice, ID and form data extraction with PaddleOCR, AWS Textract, YOLO and LayoutLM. 𝟵𝟵%+ 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 on IDs, invoices, tables and unstructured documents. 🔹 𝗛𝗲𝗮𝗹𝘁𝗵𝗰𝗮𝗿𝗲 𝗔𝗜 & 𝗘𝗠𝗥 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻: HIPAA compliant medical scribes (Whisper speech to text, speaker diarization, auto generated SOAP notes and ICD 10 codes), wound analysis computer vision, clinical RAG with PII redaction. 🔹 𝗞𝗬𝗖 & 𝗜𝗱𝗲𝗻𝘁𝗶𝘁𝘆 𝗩𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻: document localization, MRZ parsing, ArcFace biometric matching, liveness detection, real time transaction risk engines. 𝗥𝗘𝗦𝗨𝗟𝗧𝗦 𝗖𝗟𝗜𝗘𝗡𝗧𝗦 𝗣𝗔𝗜𝗗 𝗙𝗢𝗥 €𝟰𝟬𝗞 𝗽𝗲𝗿 𝘆𝗲𝗮𝗿 𝘀𝗮𝘃𝗲𝗱: automated fintech KYC pipeline (OCR plus face match plus verification agents) 𝟲𝟬% 𝗹𝗲𝘀𝘀 𝗱𝗼𝗰𝘂𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻 𝘁𝗶𝗺𝗲: AI medical scribe running in a production EMR 𝟰𝟬% 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 𝘂𝗽𝗹𝗶𝗳𝘁: hybrid GraphRAG retrieval for complex financial queries 𝟭𝟬𝘅 𝗳𝗮𝘀𝘁𝗲𝗿 𝗱𝗼𝗰𝘂𝗺𝗲𝗻𝘁 𝗽𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴: logistics and finance workflows 𝗦𝘂𝗯 𝟮𝟬𝟬𝗺𝘀 𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗹𝗮𝘁𝗲𝗻𝗰𝘆: FastAPI microservices with Redis caching 𝗧𝗘𝗖𝗛 𝗦𝗧𝗔𝗖𝗞 Python, FastAPI, Docker, AWS, GCP | LangChain, LangGraph, CrewAI, OpenAI, Claude, Hugging Face | Pinecone, FAISS, Weaviate, Neo4j, MongoDB, Redis | OpenCV, YOLO, PaddleOCR, Tesseract, LayoutLM | PyTorch, TensorFlow 𝗛𝗢𝗪 𝗜 𝗪𝗢𝗥𝗞 Fast execution with production discipline: clear milestones, regular updates, clean documented code, containerized deployment. I use modern AI dev tooling (including Claude Code) to ship in days what normally takes weeks, without cutting corners on architecture. If you want an AI system that works in production, send me an invite or message and let's scope it in a quick call. 𝗞𝗲𝘆𝘄𝗼𝗿𝗱𝘀: AI Engineer, AI Agent Developer, AI Agents, Multi Agent Systems, RAG, Retrieval Augmented Generation, Chatbot Development, LLM Integration, GPT 4o, Claude, LangChain, LangGraph, CrewAI, OCR, Document AI, Data Extraction, Computer Vision, Healthcare AI, EMR Automation, KYC Automation, Identity Verification, NLP, Python, FastAPI