Mecka
Mecka

Mecka veido fiziskā mākslīgā intelekta infrastruktūru, palīdzot robotikas un iemiesotā mākslīgā intelekta komandām pārvērst cilvēku darbības reālajā pasaulē sistēmās, kas spēj mācīties, kuras var novērtēt un kas uzticami darbojas ārpus laboratorijas. Uzņēmuma Egoverse platforma apvieno liela mēroga egocentriskas datu kopas ar telemetriju, kas aptver redzi, dziļumu, tausti, kustību un lokalizāciju.

Izmantojot izstrādātāju rīkus, API un uzņēmumu pakalpojumus, Mecka atbalsta konkrētiem uzdevumiem paredzētu robotizētu sistēmu novērtēšanu un ieviešanu. Uzņēmums sadarbojas ar aparatūras oriģinālo iekārtu ražotājiem, modeļu izstrādātājiem un komerciālajiem partneriem, nodrošinot robotikai datu un darbības slāni, pats neražojot robotus.

3D Computer Vision Engineer at Mecka (Toronto Hybrid)

Lead 3D geometry, temporal filtering, and inverse kinematics for robotics and embodied AI. Help build reliable data infrastructure and quality checks for production systems.

Apraksts

  • Develop classical 3D geometry for multi-view reconstruction, triangulation, PnP, camera calibration, undistortion, and reprojection.
  • Tune temporal filters to reduce jitter while preserving genuine motion.
  • Apply inverse kinematics and physical constraints, including rigid bone lengths.
  • Create validation checks that keep faulty frames out of delivered data.
  • Turn model outputs into clean datasets, reports, and quality metrics.
  • Maintain a data pipeline that operates at scale.
  • Create visualizers and dashboards for trajectories, overlays, and failure analysis.
  • Provide classical computer vision expertise across the organization.
  • Work with research teams and interpret model behavior without training models.

Prasības

  • At least five years of professional experience in software engineering and computer vision.
  • Write clean, efficient Python for production; C++ experience is a plus.
  • Strong command of multi-view geometry, including triangulation, PnP, camera calibration, reprojection, and coordinate transforms.
  • Experience with temporal filtering and signal processing, including low-pass and Kalman-family filters and interpolation.
  • Knowledge of inverse kinematics and rigid-body or bone-length constraints for articulated structures.
  • Experience interpreting deep model outputs and failure modes, such as those from pose and mesh regressors; model training is not required.
  • Comfort working with noisy real-world data and building quality checks that catch issues before customers do.
  • Human pose or mesh reconstruction experience is a plus.
  • Bayesian filtering, state estimation, and optimization tools such as Ceres or GTSAM are a plus.
  • Experience creating computer vision tools such as trajectory visualizers and overlay or quality dashboards is a plus.
  • Stereo, multi-view, or ego-exo capture experience is a plus.
  • Experience with lightweight temporal machine learning is a plus.

Priekšrocības

  • Equity.
  • Health and wellness benefits.
  • Additional company programs.

Saistītās vakances

Safe Fleet

Senior Software Engineer, AI Product Engineering

Safe Fleet

Build production AI features and services for Safe Fleet’s fleet safety platform. Apply large language models, machine learning, computer vision, and AI agent engineering.

Atvērt
Xero

Lead AI Workflows Engineer at Xero

Xero
1001 – 5000 Darbinieki
H1B vīzas sponsorsB2BSaaS

Lead the design and rollout of agentic AI workflows at Xero, a cloud accounting platform. Connect large language models to enterprise systems and create safe, reusable automation patterns.

Atvērt
Planitar Inc.

MLOps Engineer

Planitar Inc.

Build and operate the infrastructure used to train, deploy, and monitor Planitar’s floor-plan AI models. Support iGUIDE spatial property technology across desktop software and automated workflows.

Atvērt
ExaCare AI

Mašīnmācīšanās inženieris uzņēmumā ExaCare AI — hibrīddarbs, Toronto

ExaCare AI

Veido ražošanas ML infrastruktūru, datu plūsmas un uzraudzību, kas nodrošina ExaCare AI pēcaprūpes platformu. Palīdzi veselības aprūpes komandām pieņemt drošākus pacientu izvietošanas lēmumus, izmantojot uzticamas un mērogojamas mašīnmācīšanās sistēmas.

Atvērt