Seeking an experienced PyTorch engineer with strong knowledge of transformer internals to help implement a model-specific runtime intervention module within an existing supervisory architecture. The core architecture, evaluator, control logic, interfaces, and project structure are already implemented. This engagement focuses on implementing and validating the actuator layer that translates neutral control directives into model-specific residual-stream operations.

I have an existing iOS app, Android app, internal admin tool, website, AWS infrastructure, and integrations including Google Cloud, Twilio, and RevenueCat. I am looking for a senior US-based engineer to help streamline the existing development environment so I can safely make and preview smaller changes myself using AI tools such as Cursor, Claude Code, Codex, or similar. This is not a rebuild. The existing offshore development team will remain involved during the transition. SCOPE ~Review and document the existing codebase, database environment, and technology stack ~Confirm access to source code, AWS/GCP, Apple, Google Play, Twilio, RevenueCat, databases, and deployment systems ~Establish a clean GitHub / CI/CD workflow ~Set up AI-assisted development for the website, iOS app, Android app, internal admin tool, backend/APIs, and database-related code ~Enable safe AI-assisted changes to database queries, schemas, and migrations, with appropriate testing and safeguards before production ~Enable easy local/staging previews and testing before deployment ~Simplify and automate deployment where appropriate ~Ensure I can independently build, test, and deploy smaller changes across the website, mobile apps, admin tool, backend, and database environment ~Train me hands-on to confidently use the completed workflow VENDOR COORDINATION You will communicate directly with the existing offshore development team to understand the current environment, resolve access questions, and coordinate any necessary technical changes. You should be available for recurring meetings with the development team at: Wednesdays at 8:30 AM Eastern Fridays at 8:00 AM Eastern If either of these times is not workable, you will be expected to coordinate an alternative schedule with the offshore team that provides for two meetings per week and reasonably accommodates their working hours in India. The current vendor uses Microsoft Teams and Google Meet for meetings and communication. IDEAL CANDIDATE Strong experience with: ~AWS ~GitHub / CI/CD ~iOS and Android development/deployment ~Web applications, internal admin tools, backend APIs, and databases ~Existing production applications ~Cursor, Claude Code, Codex, or similar AI coding tools ~Technical documentation and hands-on training I prefer to work directly with a senior individual rather than an agency assigning the work to junior developers.

We're building an internal AI system that runs entirely on our own hardware (no cloud inference) against our own company data. We have a working proof-of-concept and want to get the architecture right. We need an experienced consultant to review what we've built, pressure-test our decisions, and tell us where we're wrong. This is an advisory/validation role first. We have someone doing the hands-on work; what we want is a senior second opinion to make sure we're building this the right way. NOTE: If you don't fully understand this job and you are just asking AI, don't apply. I will know if you are using AI to answer me on everything. I plan to do a thorough in person interview. What we're running today: Inference: RTX 5090 (32GB, Blackwell), Ubuntu 24.04, running llama-server (llama.cpp + CUDA) serving Gemma 4 31B-it (Q4_K_M GGUF) at a 262,144 context window. Also hosts our MCP retrieval server, PostgreSQL, and Qdrant. Embeddings: separate machine with an RTX 3060 running vLLM serving Qwen3-Embedding-4B. RAG: hybrid retrieval — Postgres full-text search + Qdrant semantic search with RRF fusion, exposed through a custom MCP server with tool-calling. Data: ingesting our own internal operational data into Postgres + Qdrant. Planned stack: LiteLLM for model routing, n8n for automation, Open WebUI for the interface, Langfuse for observability, Vault or Infisical for secrets, Keycloak/Azure AD for SSO. What we need help with: Validating our two-machine split (inference vs. embeddings) and whether our VRAM/context budget holds up under real load, specifically whether a 256K context window is real and performant on a single 32GB card or just nominal. Model selection and routing strategy: which open-weight models for which tasks, and how to structure LiteLLM routes. RAG quality: chunking, embedding dimensionality, hybrid search tuning, reranking and making retrieval actually accurate on messy real-world data. Sanity-checking our overall architecture and telling us our blind spots. You should have done: Stood up local LLM inference in production with llama.cpp/llama-server and vLLM, not just Ollama on a laptop. You understand GGUF quantization (Q4_K_M, IQ-series), KV cache, KV-cache quantization, and how context length maps to actual VRAM consumption. Real fluency in GPU sizing math given a model, a quant, and a context window, you can tell us whether it fits on a given card and what throughput to expect. Bonus if you've worked with Blackwell / sm_120a. Built production RAG vector DBs (Qdrant, pgvector), hybrid search, RRF fusion, embedding model selection, reranking, evaluation. Worked with agentic/tool-calling systems and ideally MCP servers. Know the open-weight model landscape (Gemma, Qwen, Llama, Mistral, Phi, Nemotron, Hermes) and their licenses well enough to advise. Production ops: systemd, Docker, model gateways (LiteLLM or similar), observability (Langfuse), secrets management, SSO.

Senior Computer Vision / AI Engineer — Proof-of-Concept Development AiChairWatch™ is seeking an expert-level Computer Vision / AI Engineer to develop a focused proof of concept for a patent-pending recreational seating-management technology. This is a paid, project-based independent contractor engagement. The initial objective is to establish technical feasibility and build a controlled POC—not a full commercial SaaS platform. Project Scope The POC will use computer vision and video analytics to detect people and recreational seating, establish and maintain temporary person-to-seat associations, track movement within defined spatial areas, generate state-based events, and support a lightweight operational dashboard. The engineer will be expected to help determine the appropriate technical architecture rather than simply implement a predetermined technology stack. Initial work will include: Review confidential technical requirements under NDA Evaluate technical feasibility and recommend the POC architecture Process live and/or recorded camera video Detect people and recreational seating Establish temporary person-to-seat associations Maintain associations as people move through the monitored environment Handle temporary occlusion and tracking loss Determine movement relative to configurable spatial zones Generate defined system events and state transitions Implement basic event/timing logic Create a lightweight staff-facing POC dashboard Log system events and test results Measure and document performance, limitations, and failure conditions Provide recommendations for subsequent pilot development Required Experience We are specifically looking for someone with substantial hands-on experience in real-world computer vision/video analytics. Strong experience should include several of the following: Python OpenCV YOLO or comparable object-detection frameworks Multi-object tracking Person tracking and/or re-identification Occlusion handling Spatial reasoning / regions of interest / geofencing Video-stream processing RTSP/IP camera integration Event-driven application development REST APIs/backend development Git-based source control Experience with technologies such as ByteTrack, BoT-SORT, DeepSORT, NVIDIA DeepStream, Jetson, ONNX, TensorRT, or comparable CV/edge-AI technologies is desirable but not mandatory. What We Are NOT Looking For This is not primarily a generative-AI, ChatGPT, LLM, chatbot, or prompt-engineering project. The primary technical challenge involves computer vision, persistent object/person tracking, spatial reasoning, and reliable state determination from video. Initial Development Approach The engagement is expected to begin with a paid technical feasibility and architecture milestone. The selected engineer will review the detailed requirements, evaluate representative video/test conditions, identify technical risks, recommend the detection/tracking architecture, and define measurable POC acceptance criteria. Upon successful completion of that milestone, the engagement may proceed into POC development. The initial POC will intentionally use a small controlled environment rather than attempt to build the complete commercial platform. Intellectual Property & Confidentiality AiChairWatch™ involves patent-pending technology. Detailed technical specifications and proprietary operating logic will be provided only to selected candidates after execution of a confidentiality agreement. Development work will be governed by written provisions addressing confidentiality, source code, work product, intellectual property, third-party/open-source components, and ownership/assignment of applicable development results. When Applying Please answer the following: Describe the most relevant computer-vision system you have personally designed or implemented. What experience do you have with multi-object/person tracking? How have you handled temporary occlusion, lost tracks, and re-identification? Have you processed live IP-camera/RTSP streams in a production or prototype environment? Please describe. Consider this simplified scenario: A camera observes several recreational chairs. A person occupies one chair, gets up, walks through the monitored area, is temporarily obscured by other people, and later leaves the area. At a high level, how would you maintain the person's association with the original chair and determine when that person has actually left the monitored area? What would you want to evaluate before selecting the computer-vision architecture for this POC? Please provide links to relevant GitHub repositories, demonstrations, publications, portfolio examples, or other technical work where available. Clearly identify which portions of the examples you personally designed or implemented. What is your availability during the next 60 days? Please provide your hourly rate and/or preferred structure for an initial paid technical-feasibility milestone. U.S.-based candidates preferred. AiChairWatch™ AI-Powered Recreational Seating Management Patent Pending

PhD AI/ML Engineer / Computer Scientist — Scientific ML & Physical Systems We are an early-stage technology company seeking an exceptional AI/ML engineer, computer scientist, or computational scientist to help develop the intelligence layer of a hardware-based technology platform currently in POC development. This is not an LLM, chatbot, or generative-AI project. The work involves machine learning applied to real-world physical systems and sensor data. The Role You will work directly with our hardware/firmware engineer to: Develop AI/ML models for complex physical-system data Apply physics-informed and scientific machine-learning techniques Analyze multi-channel time-series data Develop signal-processing and feature-extraction pipelines Build and evaluate anomaly and state-estimation models Develop multimodal modeling approaches Optimize models for edge/embedded deployment Integrate the intelligence layer with existing hardware and firmware Help develop the POC into a robust production architecture Ideal Background PhD strongly preferred in Machine Learning, Computer Science, Computational Physics, Electrical Engineering, Applied Mathematics, Signal Processing, or a related discipline. We are particularly interested in experience with: Physics-Informed / Scientific ML Signal Processing Time-Series ML Sensor Fusion Anomaly Detection Physical-System Modeling Embedded ML / Edge AI PyTorch / TensorFlow / JAX NumPy / SciPy C/C++ / Embedded Systems Candidates who have successfully taken ML from research into real-world physical hardware are strongly preferred. Current Stage The hardware POC and data-acquisition system are already under development. The selected candidate will work directly with our hardware/firmware engineer to develop, validate, and integrate the AI/ML layer. There is potential for a significant ongoing role for the right person. To Apply Please briefly answer: What is the most relevant scientific-ML or physical-system ML project you have personally built? What portions did you personally design and implement? What experience do you have with real-world sensor/time-series data? Have you deployed ML onto embedded or edge hardware? Please provide relevant publications, GitHub repositories, patents, or other technical work. Application details, hardware architecture, use case, datasets, system objectives, and other proprietary information will be disclosed only to selected candidates following an NDA.

We’re looking for a highly experienced AI / Full-Stack Software Developer to become our go-to technical person for a growing portfolio of live AI and SaaS products. This is NOT a role for someone who simply wants tickets assigned to them. We need someone who can own production, review other developers’ work, identify risks before they become problems, troubleshoot live issues, improve the platform, and confidently tell us whether something is ready to ship. Our products have real customers using them, which means security, reliability, speed, and accountability matter. What You'll Be Responsible For You’ll work across our existing AI/SaaS products and new development projects, including: Reviewing and approving code produced by other developers Reviewing builds before they reach production Deploying and monitoring production releases Troubleshooting and resolving live product issues Identifying security vulnerabilities and unsafe implementations Reviewing authentication, permissions, APIs and data handling Working across frontend, backend, databases, APIs and infrastructure Building and improving AI-powered features and integrations Improving monitoring, logging, alerts and internal admin tools Testing edge cases, failure scenarios and complete user journeys Managing bugs, technical debt and development priorities Improving the overall scalability and reliability of our platforms Taking ownership when something goes wrong and driving it through to resolution We want someone who naturally looks at a feature and asks: “What could go wrong when real customers start using this?” What We're Looking For Strong Full-Stack Experience You should be comfortable working across: Frontend development Backend development APIs and third-party integrations Databases Authentication and permissions Cloud infrastructure Deployment / CI/CD AI APIs, LLMs and AI integrations Monitoring and logging You don't necessarily need to be the world's best developer in every area, but you need enough experience to understand the entire product and confidently make technical decisions. Security Mindset Security is extremely important. You should be able to identify: Exposed or improperly handled data Authentication vulnerabilities Permission/access issues API security risks Unsafe implementations Secrets or credentials being exposed Infrastructure vulnerabilities Potential abuse cases Risks created by third-party integrations We want someone who thinks about security before something reaches production. Production SaaS Experience Ideally, you've previously worked on live SaaS products with paying customers. You understand that when production breaks, the answer isn't: "It worked when I tested it." You investigate the issue, determine the root cause, communicate what's happening, fix it, verify the fix and make sure it doesn't happen again. Code Review & Technical Leadership You'll often be reviewing work completed by other developers. You need to be confident enough to say: “This isn't ready for production yet, and here's why.” You'll be expected to challenge poor technical decisions, identify scalability or security concerns and maintain a high standard across the development team. How We Work This role requires high communication and accountability. We operate quickly, and priorities can change when customer or production issues arise. You'll have two short daily check-ins: Morning: priorities, releases, issues and what needs to happen that day. End of Day: what was completed, what remains outstanding, blockers and priorities carrying into tomorrow. You need to be comfortable making decisions and moving quickly without requiring constant direction. This Is NOT a Typical Freelance Role We are looking for someone who wants to become a core part of our company, not someone managing ten unrelated Upwork projects simultaneously. Our company needs to be your primary development commitment and priority. If something important happens with a live product, we need our technical lead to be available, responsive and willing to take ownership. There is significant opportunity for this role to become a long-term position as our software and AI products continue to grow. The Person We're Looking For You'll probably be a great fit if you're: Highly proactive Security conscious Quality obsessed Fast and decisive Extremely organized Calm when something breaks Comfortable challenging other developers Strong at troubleshooting Commercially aware Comfortable taking responsibility Constantly looking for ways to improve a platform You understand that technical decisions don't just affect code. They affect customers, retention, reputation, revenue and risk. To Apply Please answer the following: 1. Tell us about a live SaaS product you've personally been responsible for in production. 2. Describe a serious production issue you've dealt with. What happened, how did you diagnose it, and what did you do to prevent it happening again? 3. Give us an example of a security vulnerability or risk you've identified before something went live. 4. What's your process for reviewing another developer's code before approving it for production? 5. What experience do you have building or maintaining AI-powered software? 6. What is your current development stack? 7. How many other clients/projects would you continue working with if you took this position? 8. Are you comfortable with two short daily accountability meetings? 9. What hours/time zone do you normally work? 10. What is your availability if a critical production issue occurs? Please start your application with “PRODUCTION” so we know you've read the entire posting. We're looking for someone who wants to take ownership of a platform, not simply complete development tasks.

Vondy builds and maintains custom software and AI applications for mid-market businesses. Our work includes document processing, internal tools, customer-facing applications, and more. We are a team of Forward Deployed Engineers who can move between understanding a customer workflow, rapidly prototyping a solution, and shipping a reliable production application. Responsibilities - Translate operational workflows into clear software requirements - Build full-stack web applications and internal tools - Integrate APIs, databases, documents, and existing business systems - Implement LLM-powered workflows, agents, and structured outputs - Work directly with Vondy’s founders and occasionally with customer teams - Test, document, deploy, and maintain production software Initial Paid Onboarding - Official Anthropic Developer or Architect Foundations preparation - A qualifying Anthropic certification exam, with the exam cost covered by Vondy Requirements - At least three years of professional software-development experience - Strong TypeScript/JavaScript, Node.js, or Python experience - Experience shipping full-stack applications to production - Familiarity with LLM APIs, tool use, agents, or document-processing systems - Strong product judgment and ability to work from incomplete business requirements - Clear written and spoken English - Reliable availability and interest in an ongoing project-based relationship - Ability to complete the initial onboarding promptly once access is available Nice To Have - Experience with Claude, Claude Code, MCP, or Anthropic’s API Please Include - A production application you personally helped build - Your experience with AI or LLM-powered applications - Confirmation that you are willing to complete the paid Anthropic certification process

I’m looking for a developer to help build a lightweight AI prototype using OpenAI or Anthropic APIs. This is NOT a full product build. This is a focused prototype to test a specific idea. Project Goal: Build a simple Python-based system that: Runs the same LLM task multiple times. Captures outputs and any intermediate state (memory/logs). Compares differences between runs. Classifies differences into simple categories: Stable Boundary Violation What This Means Think: •Run the same prompt 5–10 times. •Log results. •Detect where outputs or stored data differ. •Label those differences. That is it. Technical Requirements Must have: •Python •Experience with OpenAI API or Anthropic API •Ability to build simple, clean scripts (no over-engineering) Nice to have: •LangChain or similar frameworks. •Streamlit (for simple UI/dashboard). •Experience with logging or comparing outputs. Important Constraints This should be: •Lightweight. •fast to build. •easy to understand. Please DO NOT: •Design complex architectures. •build full systems. •over-engineer. Deliverables •Python script or small app. •Ability to run repeated LLM tasks. •Stored logs of runs (JSON or similar). •Basic comparison logic between runs. •Simple classification output. Timeline •3–7 days initial build •Max 1–2 weeks total Engagement Style •Fixed-price or hourly (open to discussion) •Will start with a small paid test task before full project Screening Question (Required) Please answer this: If you needed to run the same LLM task multiple times and compare outputs/state between runs, how would you build it quickly? Who This Is For Ideal candidate: •Builds fast prototypes. •Comfortable with LLM APIs. •Prefers simple solutions over complex systems.

CONTACTING OUTSIDE OF UPWORK WILL RESULT IN AUTOMATIC DISQUALIFICATION Description: We're an agency staffing infrastructure and MLOps talent across a portfolio of active platform builds for confidential enterprise clients. Work includes provisioning environments inside client-owned cloud infrastructure under their security review process, and building CI/CD and observability for both application and production ML systems. Immediate need — looking to onboard within the next 5-7 business days. This role is typically front-loaded (environment setup at kickoff) then lighter mid-build, ramping again ahead of go-live — hours will flex accordingly rather than staying flat throughout. What you'll do: Stand up and manage CI/CD pipelines for both application and ML model deployment Implement model registry and shadow/champion-challenger release patterns with automated rollback Build observability — logging, monitoring, alerting — across application and ML layers Provision and manage environments within client-owned cloud infrastructure per their security review requirements Support drift monitoring and retraining triggers for production ML systems Required: Strong DevOps/infrastructure-as-code background (Terraform, CloudFormation, or similar) CI/CD pipeline design experience, including for ML model deployment specifically (MLOps) Cloud platform experience (AWS, GCP, or Azure), including provisioning within a client-owned/managed environment Available to start within the next week Nice-to-have: ML monitoring/observability tooling (drift detection, model registries such as MLflow), security-review or compliance-adjacent infrastructure experience

CONTACTING OUTSIDE OF UPWORK WILL RESULT IN AUTOMATIC DISQUALIFICATION Description: We're an agency staffing infrastructure and MLOps talent across a portfolio of active platform builds for confidential enterprise clients. Work includes provisioning environments inside client-owned cloud infrastructure under their security review process, and building CI/CD and observability for both application and production ML systems. Immediate need — looking to onboard within the next 5-7 business days. This role is typically front-loaded (environment setup at kickoff) then lighter mid-build, ramping again ahead of go-live — hours will flex accordingly rather than staying flat throughout. What you'll do: Stand up and manage CI/CD pipelines for both application and ML model deployment Implement model registry and shadow/champion-challenger release patterns with automated rollback Build observability — logging, monitoring, alerting — across application and ML layers Provision and manage environments within client-owned cloud infrastructure per their security review requirements Support drift monitoring and retraining triggers for production ML systems Required: Strong DevOps/infrastructure-as-code background (Terraform, CloudFormation, or similar) CI/CD pipeline design experience, including for ML model deployment specifically (MLOps) Cloud platform experience (AWS, GCP, or Azure), including provisioning within a client-owned/managed environment Available to start within the next week Nice-to-have: ML monitoring/observability tooling (drift detection, model registries such as MLflow), security-review or compliance-adjacent infrastructure experience