Computer Vision Engineer

Role Summary

We are seeking a Junior-Mid Level Computer Vision Engineer to design and implement real-time vision pipelines that process camera input to extract meaningful information from user environments. A core responsibility is building robust systems that capture and process images from mirrors and reflective surfaces—a challenging problem that requires careful handling of camera geometry, reflection properties, and image preprocessing.

This role is ideal for someone with 3–4 years of hands-on computer vision experience who is ready to take ownership of core perception components.

Key Responsibilities

1. Multi-Camera System Design & Integration

• Design and implement multi-camera capture systems (2–3 synchronized cameras at 30 FPS).

• Build camera synchronization and frame buffering mechanisms to ensure temporal alignment.

• Implement intrinsic and extrinsic camera calibration pipelines using standard patterns (checkerboards) and robust optimization techniques.

• Create diagnostic tools to validate calibration quality, reprojection error, and baseline consistency.

• Handle camera configuration across different deployment scenarios (phones, tablets, dedicated hardware).

2. Image Processing & Preprocessing

• Implement image preprocessing pipelines tailored to real-world capture conditions.

• Lens distortion correction (radial and tangential).

• Histogram equalization and contrast normalization for varying lighting.

• Temporal filtering and smoothing to reduce noise and jitter.

• Handle mirror and reflective surface processing.

• Detecting reflections and mirrored regions in images.

• Correcting for mirror geometry (horizontal and vertical flipping, coordinate transformations).

• Extracting valid regions from mirror-captured views while handling edge artifacts.

• Implementing fallback mechanisms for single-mirror vs. multi-mirror setups.

• Develop robust preprocessing that adapts to different environments (gym, home, outdoor).

3. 2D Object/Keypoint Detection Pipeline

• Integrate pre-trained 2D detection models (e.g., HRNet, YOLOv8, OpenPose) into production pipelines.

• Implement keypoint post-processing and confidence filtering.

• Filtering low-confidence detections.

• Handling partially occluded or out-of-frame detections.

• Temporal consistency checks across frames.

• Build validation frameworks to assess detection quality on different body types, clothing, and capture angles.

• Optimize detection inference for real-time performance (<100ms end-to-end).

4. 3D Triangulation & Reconstruction

• Implement 3D point triangulation from 2D keypoints using calibrated multi-camera systems.

• Direct Linear Transform (DLT) for robust 3D lifting.

• Handling degenerate cases (parallel camera views, poor baseline).

• Reprojection error computation for reconstruction quality assessment.

• Build 3D reconstruction validation frameworks.

• Accuracy metrics (reprojection error, 3D distance validation).

• Error analysis and outlier detection.

• Integrate multi-frame temporal filtering to smooth 3D trajectories and reduce jitter.

5. Real-Time Performance Optimization

• Profile vision pipelines to identify bottlenecks (CPU, memory, GPU utilization).

• Optimize critical components for sub-100ms latency.

• Algorithm-level optimizations (vectorization, parallelization).

• Memory-efficient data structures and buffer management.

• GPU utilization where applicable (CUDA, OpenCL basics).

• Implement efficient frame streaming and buffering mechanisms.

• Create performance monitoring and logging to detect degradation.

6. Testing, Validation & Documentation

• Design and implement comprehensive test suites for all vision components.

• Unit tests for individual algorithms (calibration, triangulation, preprocessing).

• Integration tests for end-to-end pipelines.

• Regression tests to catch performance degradation.

• Validate system accuracy on diverse scenarios.

• Different lighting conditions and environments.

• Varying camera positions and baselines.

• Challenging edge cases (occlusion, mirror artifacts, reflections).

• Maintain clear documentation of.

• Algorithm implementations and assumptions.

• Calibration procedures and validation protocols.

• Known limitations and failure modes.

• Performance benchmarks and optimization decisions.

7. Cross-Functional Collaboration

• Partner with the Senior CV Engineer on architectural reviews and algorithm validation.

• Collaborate with the Mobile ML Engineer on model integration and latency requirements.

• Work with Full-Stack Engineer on API contracts and data formats for vision outputs.

• Participate in technical design reviews and share findings with the team.

Required Skills & Experience (Junior-Mid Level)

Educational Background

• Bachelor's degree in Computer Science, Engineering, Mathematics, Physics, or related field.

• Master's degree (MS) is a plus but not required.

• Strong foundation in linear algebra, geometry, calculus, and numerical methods.

Core Programming & Libraries

• 3–4 years hands-on experience in computer vision (production, research, or substantial projects).

• Expert-level Python (3+ years).

• NumPy, SciPy for numerical computing.

• Matplotlib, Seaborn for visualization.

• Scikit-image for image processing fundamentals.

• Proficient with OpenCV (3+ years).

• Camera calibration (cv2.calibrateCamera, cv2.undistort).

• Stereo rectification and triangulation (cv2.triangulatePoints, cv2.stereoRectify).

• Feature detection and matching (cv2.SIFT, cv2.ORB).

• Image transformations and geometric operations.

• Video I/O and frame capture.

• Comfortable with C++ (1+ year).

• Ability to read and modify performance-critical vision code.

• Understanding of GPU/CUDA basics.

• Version control: Git workflows, collaborative development.

Computer Vision & 3D Geometry (Proficient)

Essential Knowledge

• Camera models: Pinhole camera geometry, intrinsic matrices (K), extrinsic matrices (R, t), perspective projection.

• Lens distortion: Radial and tangential distortion models, correction procedures.

• Epipolar geometry: Fundamental matrix computation, stereo rectification, baseline and depth relationships.

• 3D Triangulation: Direct Linear Transform (DLT), triangulation from multiple views, handling degenerate cases.

• Multi-view geometry: Understanding homographies, coordinate transformations, and projection relationships.

Mirror & Reflection Processing (Specific to Role)

• Understanding of mirror optics: plane mirror reflection properties, coordinate transformations for reflected views.

• Techniques for detecting mirrors in images (edge detection, smooth surface characteristics).

• Mirror-corrected triangulation: Handling 3D points reconstructed from direct and mirrored views, coordinate frame conversions.

• Robustness to reflection artifacts (specular highlights, partial reflections, mirror edges).

Hands-on Experience with

• Camera calibration using checkerboard or similar patterns.

• Stereo vision and depth computation.

• Image warping and geometric alignment.

• Pose estimation model integration and post-processing.

3D Geometry & Spatial Reasoning

• 3D coordinate systems and transformations (rigid transformations, rotation matrices).

• Vector and matrix operations in 3D space.

• Rotation representations: Rotation matrices, Euler angles, quaternions (conceptual and computational).

• Triangulation mathematics: 2D-to-3D lifting, least-squares solutions, numerical stability.

• Understanding of coordinate frame conversions: world, camera, and local body frames.

Real-Time Systems & Performance

• Experience optimizing code for latency targets (frame-rate constraints).

• Comfortable profiling Python and C++ code (timing, memory usage, CPU/GPU utilization).

• Understanding of GPU vs. CPU trade-offs and basic GPU programming concepts.

• Frame synchronization and temporal alignment in multi-camera systems.

• Experience with video pipelines and streaming data handling.

Mathematical Knowledge (Proficient)

• Linear algebra: Matrix operations, decomposition basics (SVD, QR), eigenvalues, homogeneous coordinates.

• Geometry: Dot products, cross products, projective geometry, 3D transformations.

• Numerical methods: Least-squares fitting, handling ill-conditioned matrices, numerical precision.

• Calculus: Partial derivatives, optimization basics, understanding of gradients in geometric problems.

Communication & Documentation

• Ability to write clear, modular, well-documented code.

• Technical writing: Document algorithms, procedures, calibration protocols, and design decisions.

• Comfortable explaining complex geometric concepts to other engineers and product teams.

• Strong code review practices and receptiveness to feedback.

Preferred Skills & Experience

• Production computer vision systems: Experience deploying CV models or pipelines in real-world applications.

• Multi-camera systems: Prior work with stereo rigs, calibration, or synchronized capture.

• Real-time CV: Performance-critical vision applications (robotics, AR/VR, autonomous systems, sports analytics).

• Specialized domains: Experience with reflective surfaces, specular highlights, or challenging lighting environments.

• Video processing: Working with video streams, temporal coherence, frame extraction.

• Deep learning integration: Experience integrating pre-trained CV models in production pipelines.

• Optimization & profiling: GPU optimization (CUDA), parallel processing, memory optimization.

• Game engine experience: Graphics concepts, 3D visualization, shader programming basics.

• Research or publications: Familiarity with computer vision papers from CVPR, ICCV, ECCV, or similar venues.

• Open-source contributions: GitHub repositories demonstrating CV project work.

What You'll Gain

• Hands-on computer vision expertise on complex, real-world problems (multi-camera systems, mirror processing, real-time optimization).

• Mentorship from a senior engineer: Work directly with an experienced CV specialist who will guide your technical growth.

• Production impact: Your vision systems power intelligent products used across different domains.

• Deep technical learning: Real-time optimization, multi-view geometry, camera systems, and robustness engineering applied in practice.

• Autonomy with support: Own specific perception components while receiving guidance from senior engineer.

• Technical leadership trajectory: Clear path to senior CV engineer roles as you grow.

• Dual ecosystem exposure: Engage with both Indian tech ecosystem and international CV research community.

• Hybrid working opportunities: Post MVP phase (Month 6 onwards), flexibility for remote collaboration as per team needs.

• Patent involvement: Potential contribution to Nutpaa's computer vision patents.

Work Arrangement

• Commitment: Full-time, 45–50 hours/week (standard startup pace)

• Location: Madurai, Tamil Nadu / CV lab within Tamil Nadu.

• Post-MVP: Hybrid working opportunities from Month 6 onwards with flexibility for remote collaboration

Organizational & Cultural Expectations

• Maintain technical rigor in algorithm implementation and testing.

• Share learnings through code reviews, documentation, and team discussions.

• Provide and receive feedback with clarity and respect.

• Uphold Nutpaa's values: Engineering Excellence, Long-Termism, Open Evolution, and Peer-Driven Collaboration.

• Take initiative in learning and solving problems, with support from senior engineer.

• Challenge assumptions constructively and help improve team practices.

• Maintain confidentiality of proprietary algorithms and data.

• Contribute to technical discussions and code quality improvements.

Application Process

Please email to careers@nutpaa.ai with:

1. Resume (highlighting computer vision projects, hands-on experience; include academic credentials and any degrees/certifications)

2. Portfolio

• GitHub repos showing computer vision projects or implementations

• Links to any published work, blogs, or technical writing

• Examples of OpenCV-based projects, image processing work, or camera calibration

• Evidence of real-time or performance-critical CV work

• Examples of multi-camera or 3D reconstruction projects

3. Brief statement (~200 words)

• Why you're interested in real-time computer vision systems and Nutpaa's mission

• One example of a complex computer vision problem you solved (include technical details and learnings)

• Your preferred learning style and how you approach technical growth

4. Optional but valuable

• Technical writing sample (blog post, project documentation, algorithm explanation)

• Links to any open-source contributions or personal projects

• Examples of multi-camera, mirror processing, or specialized vision work

Email Subject: Junior-Mid Level Computer Vision Engineer – [Your Name]

Equal Opportunity Statement

Nutpaa is an equal opportunity provider. We are committed to building a diverse and inclusive team, and we do not discriminate based on race, religion, color, national origin, gender, gender identity or expression, sexual orientation, age, marital status, veteran status, or disability status.