Lead - Computer Vision & AI Architect

Eonix Partners Llp; Bengaluru; On-site; INR 6,000,000 - 11,000,000; Full time

Job summary

Eonix Partners Llp in Bengaluru is seeking a Lead Computer Vision & AI Architect to own the technical direction of our perception stack while staying hands-on. You will guide the design of our real-time movement-perception system across cameras, 3D pose, and edge deployment, connecting capture, edge runtime, mobile app, and cloud.

This hybrid role blends technical leadership with product ownership. You will set the roadmap, translate goals into module requirements, and mentor a small senior team

Qualifications

• 7–10 years in computer vision, AI, or ML
• Experience owning technical direction and architecture
• Delivery of real-time systems with strict latency targets

Responsibilities

• Set and own architecture decisions across 3D pose, multi-camera geometry, body/mesh modeling, form evaluation, and edge deployment.
• Translate product goals into module-level requirements and acceptance criteria.
• Plan and run sprints, manage milestones and KPIs (latency, frame rate).
• Mentor CV and AI/ML engineers; ensure code quality and testing.

Skills

Technical leadership

Hands-on coding

Architectural design

Cross-functional

Real-time systems

CV/ML expertise

Education

Master's or PhD in CS/ML

Tools

Python

OpenCV

C++

NumPy/SciPy

Job description

We are looking for a Lead Computer Vision & AI Architect to own the technical direction of our core perception stack while staying deeply hands-on. You will lead the design of our real-time movement-perception system multi-camera geometry, 3D pose and mesh estimation, personalized body modeling, live form evaluation, Vision-Language coaching models, and edge deployment and you will also carry end-to-end responsibility for how these pieces fit together as one coherent system across capture, edge runtime, mobile app, and cloud.

This is a hybrid technical-lead and product-ownership role, not a pure people-management position. You will set the perception roadmap, own module-level requirements and acceptance criteria, plan and run execution, and act as the primary technical bridge across engineering, design, hardware, and partner integrations. You will mentor a small, senior team through design and code review — formal people management is not the core of the role today, but technical leadership and delivery ownership are.

Key Responsibilities

1. Technical Leadership & Execution Ownership

• Technical Direction: Set and own architecture decisions across 3D pose, multi-camera geometry, body/mesh modeling, form evaluation, and edge deployment — and be accountable for their trade-offs.
• Product Ownership: Translate product goals into clear module-level requirements, scope, priorities, and acceptance criteria for the perception program; decide what ships and what defers.
• Execution Management: Plan and run sprints, sequence cross-module work, and manage delivery against milestones and measurable KPIs (latency, frame rate, calibration repeatability, demo reliability).
• Engineering Practice: Foster clean coding, thoughtful design and code reviews, automated testing, clear documentation, and reproducible experiment tracking. Mentor CV and AI/ML engineers technically.

2. System Architecture & Cross-Functional Coordination

• System Coherence: Own end-to-end architecture coherence across capture, perception, edge runtime, mobile app, and cloud — not just the CV core, but how the whole movement-intelligence loop connects.
• Interfaces & Dependencies: Define and manage cross-module interface contracts and dependencies so parallel workstreams integrate cleanly.
• Cross-Functional Bridge: Coordinate day-to-day across full-stack engineering, UI/UX, CV and ML/CV engineers, and — as the team grows — edge/embedded and hardware/firmware engineers, plus partner integrations.
• Performance Budgets: Establish and enforce real-time latency, frame-rate, memory, and thermal budgets across edge hardware, smart glasses, Linux workstations, and NPUs — from current real-time demo targets (sub-second feedback at 10–15 FPS on Jetson-class simulation) toward productization targets (sub-200ms feedback at 30–60 FPS).

3. Hands-On Computer Vision & 3D Geometry

• Multi-View Geometry & 3D Pose: Lead hands-on development of camera calibration, multi-camera synchronization, epipolar geometry, 3D joint triangulation, and temporal smoothing across a multi-camera capture rig.
• Personalized Body Modeling & Form Evaluation: Build pose-to-mesh alignment against a user-specific skeletal/mesh model and a live form-comparison engine that turns movement variance into timely, trustworthy corrective cues.

4. Multimodal AI & Edge Deployment

• Multimodal AI: Guide evaluation and fine-tuning (LoRA/QLoRA) of Vision-Language Models (e.g. LLaVA, Qwen-VL) for movement interpretation and coaching dialogue.
• Edge Optimization: Oversee quantization (INT8/FP16) and export to ONNX, TensorRT, TFLite, or CoreML, and real-time inference on Jetson-class edge and NPU targets.

5. R&D and Intellectual Property

• Patent-Protected Innovation: Help build and document patentable core technology in mirror-aware tracking, personalized 3D motion reconstruction, and real-time form evaluation.

What We Are Looking For

Experience & Leadership

• Background: 7–10 years in computer vision, AI, or ML engineering, with a track record of owning technical direction and architecture on complex, real-time systems.
• Hands-On + Coordination: Demonstrated ability to stay deeply hands-on in code while also leading architectural reviews, driving execution, and coordinating delivery across cross-functional teams.
• Delivery: Proven record of shipping complex engineering projects against deadlines, with rigorous benchmarking and clear technical decision-making.

Computer Vision & 3D Geometry

• Core Math: Strong grasp of 3D geometry (rigid-body transforms, quaternions, rotation matrices) and projective geometry (calibration, bundle adjustment, triangulation).
• Pose Systems: Practical experience with 2D/3D pose models (e.g. ViTPose, HRNet, RTMPose, MediaPipe) and multi-camera setups.
• Toolkit: High proficiency in Python (NumPy, SciPy), OpenCV, and performance-oriented C++.

AI/ML & Multimodal AI

• Frameworks: Strong experience with PyTorch and Hugging Face Transformers.
• Multimodal Models: Hands-on work evaluating and fine-tuning Vision-Language Models (VLMs).
• Agentic AI: Familiarity with agent systems, tool calling, or Retrieval-Augmented Generation (RAG) is a plus.

Edge & Deployment

• Hardware Targets: Experience delivering real-time models to mobile SoCs, NPUs, smart glasses, or embedded hardware.
• Optimization Tools: Hands-on experience with TensorRT, ONNX Runtime, TFLite, or CoreML.

Nice to Have

• Education: Master's or Ph.D. in Computer Science, Computer Vision, Machine Learning, or a related field.
• Publications or Patents: Work published at CVPR, ICCV, ECCV, or NeurIPS, or granted patents.
• Domain Experience: Familiarity with biomechanics, fitness technology, sports analytics, or AR/VR.