New opportunity

Mantra Softech - Senior Edge AI Engineer - iOS & Mac OS

Mantra Softech · Ahmedabad, Bangalore, India

About this role

Mantra Softech seeks a Senior Edge AI Engineer to build and optimize production-grade biometric SDKs and deploy AI/ML models on iOS and macOS. The posting lists Ahmedabad and Bangalore as locations and an India applicant-location requirement.

Responsibilities

Develop reusable SDKs using XCFramework, Swift Package Manager and CocoaPods for face, fingerprint, iris and palm biometrics, Document AI, OCR and liveness detection. Design SDK architecture, public APIs, documentation, sample applications and version management.
Deploy and optimize models with Core ML, ONNX Runtime, TensorFlow Lite and PyTorch Mobile, using Apple Neural Engine, Metal and Accelerate Framework. Build camera and image-processing pipelines with AVFoundation, Vision Framework, Core Image, OpenCV and real-time video processing.
Improve latency, memory use, battery consumption, multithreading and application performance. Implement security using Keychain, Secure Enclave, CryptoKit, certificate pinning, jailbreak detection and anti-tampering.
Develop automated tests, benchmarks, profiling and CI/CD pipelines for SDK releases. Collaborate with AI, embedded, SDK and platform teams to deliver production-ready biometric solutions and maintain cross-platform feature parity.

Qualifications and listed skills: The posting specifies 6–10 years of experience and lists a bachelor's degree in its structured requirements. Listed skills include Swift, Objective-C, C++, OpenCV, SDK development, macOS, iOS and Edge AI.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

Edge AIiOSmacOSSwiftObjective-CC++XCFrameworkSwift Package ManagerCocoaPodsCore MLONNX RuntimeTensorFlow LitePyTorch MobileApple Neural EngineMetalAccelerate FrameworkAVFoundationVision FrameworkCore ImageOpenCVDocument AIOCRBiometricsImage processingSDK developmentAPI designMultithreadingKeychainSecure EnclaveCryptoKit

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Reusable biometric SDKs
The role develops iOS and macOS SDKs for face, fingerprint, iris and palm biometrics, plus Document AI, OCR and liveness detection. Prepare to explain architecture and packaging decisions for XCFramework, Swift Package Manager and CocoaPods.
Stable developer-facing interfaces
The posting calls for public APIs, documentation, sample applications and version management. Prepare to discuss how you would make an SDK usable while managing changes across releases.
On-device model deployment
The role deploys and optimizes models through Core ML, ONNX Runtime, TensorFlow Lite and PyTorch Mobile, using Apple Neural Engine, Metal and Accelerate. Prepare to explain how you would select a runtime and measure its results.
Real-time pipeline performance
The role builds camera and image-processing pipelines and optimizes latency, memory, battery use and multithreading. Prepare a concrete approach to profiling video processing and diagnosing a bottleneck.
Secure, tested releases
The posting specifies Keychain, Secure Enclave, CryptoKit, certificate pinning, jailbreak detection and anti-tampering, alongside automated tests, benchmarks, profiling and CI/CD. Prepare to discuss how security checks and performance evidence fit into an SDK release.

Likely challenges to think through

Portability versus acceleration
Possible challenge, not a reported problem: a model may be easier to share across runtimes than to optimize for Apple hardware. Prepare to describe a comparison using accuracy, latency, memory and device-level benchmarks before choosing a deployment path.
Responsive capture versus resource use
Possible challenge: real-time camera processing may improve responsiveness at the cost of battery or memory. Prepare to explain how you would profile pipeline stages, set performance targets and test trade-offs on supported Apple devices.
Feature parity versus platform-specific design
Possible challenge: coordinating with AI, embedded, SDK and platform teams on cross-platform parity may conflict with an API tailored to iOS and macOS. Prepare to describe how you would define shared behavior, document platform differences and test releases for regressions.

Practice your answers

Technical practice

How would you package a reusable biometric SDK for both iOS and macOS?

What to coverSuggestion: Compare XCFramework, Swift Package Manager and CocoaPods in terms of distribution, dependencies and integration effort. Explain how you would verify that sample applications can consume each release.

How would you design public APIs and versioning for an SDK that supports several biometric capabilities?

What to coverSuggestion: Discuss stable interfaces for face, fingerprint, iris, palm and liveness features without exposing unnecessary implementation details. Address documentation, compatibility and how an API change would be communicated.

How would you choose and optimize a model runtime for on-device inference on Apple platforms?

What to coverSuggestion: Compare Core ML, ONNX Runtime, TensorFlow Lite and PyTorch Mobile based on model compatibility, measured performance and integration cost. Explain when Apple Neural Engine, Metal or Accelerate might help, while treating device-specific results as something to benchmark.

How would you build a real-time camera-to-inference pipeline?

What to coverSuggestion: Walk through capture, image processing and inference using relevant options such as AVFoundation, Vision, Core Image and OpenCV. Discuss frame handling and the trade-off between image quality and end-to-end latency.

How would you evaluate a change to a biometric, OCR or liveness feature before shipping it in an SDK?

What to coverSuggestion: Describe representative input tests, correctness checks and comparisons against the current release for the feature at hand. Include latency and failure cases alongside output quality rather than assuming one metric establishes production readiness.

How would you diagnose an SDK that meets inference targets but slows its host application?

What to coverSuggestion: Separate model execution from camera processing, memory allocation and multithreading effects when profiling. Explain how you would weigh latency improvements against memory use, battery consumption and overall application performance.

How would you protect sensitive data and SDK communications on iOS and macOS?

What to coverSuggestion: Explain the distinct roles of Keychain, Secure Enclave, CryptoKit and certificate pinning where applicable. Discuss what jailbreak detection and anti-tampering can mitigate, without treating either as an absolute guarantee.

What would an automated test and release strategy for a production biometric SDK include?

What to coverSuggestion: Cover API and integration tests, repeatable benchmarks, profiling and CI/CD checks before release. Explain how you would detect performance regressions as well as functional failures.

How would you manage Swift, Objective-C and C++ boundaries in a reusable Apple-platform SDK?

What to coverSuggestion: Discuss where interfaces should sit and how you would handle data ownership, error propagation and maintainability across language boundaries. Relate the choices to stable public APIs and integration by SDK consumers.

How would you maintain feature parity between iOS and macOS without ignoring platform differences?

What to coverSuggestion: Identify which SDK behavior and public contracts should remain consistent, then test platform-specific camera, processing and model-execution paths separately. Explain how you would assess and document a justified difference.

Behavioural practice

Tell me about a time you coordinated model integration across engineering teams.

What to coverSuggestion: Set out the situation and each team's responsibilities, then describe the actions you took to resolve an integration issue with AI, embedded, SDK or platform counterparts. Finish with the observable result and what you learned.

Tell me about a time you improved performance without compromising another important constraint.

What to coverSuggestion: Describe the starting latency, memory, battery or application-performance problem and how you measured it. Explain your action, the trade-off you made and the measured result.

Tell me about a time testing or profiling changed a release decision.

What to coverSuggestion: Describe the release situation, the evidence you found and your role in deciding what to fix or defer. State the result, including how the issue was verified or communicated.

Tell me about a time you had to change a public API used by other developers.

What to coverSuggestion: Explain the original constraint, the options you considered and how you handled compatibility, documentation or versioning. Describe the outcome for developers using the API.

Tell me about a time you handled a feature discrepancy across platforms.

What to coverSuggestion: Describe the expected behavior, the difference you found and how you worked with others to decide whether parity was feasible. Explain your action and the resulting behavior or documented limitation.

Sharpen your knowledge

Topics to prepare

SDK architecture and distribution
The posting calls for reusable iOS/macOS SDKs, public APIs, documentation, sample applications and version management using XCFramework, Swift Package Manager and CocoaPods. For revision, sketch an SDK boundary and explain how an API change would affect consumers and releases.
On-device model deployment
The role names Core ML, ONNX Runtime, TensorFlow Lite and PyTorch Mobile, alongside Apple Neural Engine, Metal and Accelerate. For revision, compare how you would select and validate a runtime and acceleration path for a biometric model; this comparison is a preparation exercise, not a stated selection process.
Biometric image pipelines
The posting covers face, fingerprint, iris and palm biometrics, Document AI, OCR and liveness detection, with real-time camera processing. For revision, diagram capture, image processing and model inference stages, noting where latency or image quality could affect results.
Performance and concurrency
The SDKs must be optimized for latency, memory, battery use, multithreading and application performance. For revision, explain a measurement-led approach to finding a video-pipeline bottleneck and distinguishing computation from memory or threading costs.
SDK security
The posting names Keychain, Secure Enclave, CryptoKit, certificate pinning, jailbreak detection and anti-tampering. For revision, map which SDK assets and operations each measure could protect, and discuss its limitations rather than treating any measure as a guarantee.
Release quality and cross-platform parity
The role includes automated tests, benchmarks, profiling, CI/CD releases and collaboration across AI, embedded, SDK and platform teams. For revision, outline how you would check behavior and performance on both iOS and macOS before an SDK release.

Skills to revise

Package and expose an SDK
The posting specifies XCFramework, Swift Package Manager and CocoaPods, and lists Swift, Objective-C and C++. As practice, package a small Swift API, write a minimal consumer example and check that its public interface and versioning are clear; this is a suggested exercise, not a required deliverable.
Run an on-device inference comparison
Model deployment and optimization are explicit responsibilities. As practice, run the same suitable sample model through two named runtimes where feasible, record output agreement and latency, and explain any trade-offs without assuming the employer uses that model.
Build a camera-to-inference prototype
The posting names AVFoundation, Vision Framework, Core Image and OpenCV for image and real-time video processing. As practice, capture frames, apply a simple preprocessing step and pass them to an inference interface while observing frame rate.
Profile and improve a bottleneck
Latency, memory, battery use and multithreading are explicit optimization targets. As practice, benchmark a small processing pipeline, change one implementation choice, then compare measurements and explain the trade-off.
Implement a focused security example
Secure SDK implementation is an explicit responsibility. As practice, protect a sample secret with Keychain or demonstrate a CryptoKit operation, then describe what it protects and what remains outside its scope.
Automate an SDK release check
The posting calls for automated testing, benchmarking, profiling and CI/CD for SDK releases. As practice, create a small test and benchmark suite for an SDK API and define checks that would catch an iOS/macOS behavior difference before release.

Your preparation checklist

Select a relevant project walkthrough
Prepare a concise account of an iOS or macOS SDK or on-device AI project, if available. Show your design decisions, your contribution, measured results and lessons learned; the role centers on production-grade biometric SDKs.
Map evidence to the listed skills
Create a one-page evidence sheet for relevant work in Swift, Objective-C, C++, OpenCV, SDK development and Edge AI. Include concrete examples and dates rather than claims of proficiency; the posting lists 6–10 years of experience.
Sketch a reusable SDK design
Prepare a design exercise showing public APIs, documentation, a sample app and versioning for an iOS/macOS SDK. Explain how you would package it with XCFramework, Swift Package Manager or CocoaPods.
Plan an on-device model comparison
Outline how you would evaluate a biometric or OCR model using the listed deployment options, such as Core ML and ONNX Runtime. Propose measurements for latency, memory and battery use, and explain when Apple Neural Engine, Metal or Accelerate might help; treat this as a proposed approach, not a company requirement.
Diagram a real-time capture pipeline
Draw a camera-to-result flow using AVFoundation and relevant image-processing tools such as Vision, Core Image or OpenCV. Be ready to discuss frame handling, multithreading and performance trade-offs.
Prepare a security discussion
Draft a threat model for a biometric SDK and explain possible uses and limitations of the posting’s listed controls: Keychain, Secure Enclave, CryptoKit, certificate pinning, jailbreak detection and anti-tampering.
Show how you would validate a release
Prepare a proposed test and benchmark plan covering SDK APIs, model outputs, performance profiling and CI/CD release checks. Include how you would coordinate with AI, embedded, SDK and platform teams on cross-platform feature parity.
Bring focused questions for the employer
Ask which biometric or Document AI use cases are priorities, which Apple platforms and model runtimes matter most, how SDK performance is measured, and how cross-platform parity is defined. These are clarification questions, not assumptions about Mantra Softech’s current process.

Availability last checked: 5 Nov 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…