Graduate Engineering Trainee

Graduate Engineering Trainee (AI)

Title: Edge/Embedded AI Engineer (GET)

Function: AI
• Computer Vision
• Embedded Systems

Location: HSR Layout, Bengaluru (on‑site; occasional field/device data collection)

Employment Type: Full‑time trainee (6 months program)

Experience: 0-1 year | Qualification: B.E/B.Tech (CS/IS/AI) with strong programming fundamentals

Compensation :

• During the 6-month GET term, you'll receive a stipend up to ₹25,000/month (role fit & performance based).
• On successful completion and confirmation, the role converts to ₹7,50,000 CTC per annum.
• Unlimited coffee and conversation with Founders is on the house

About Mankomb

Mankomb Technologies builds AI‑native appliances for sustainable urban living.

Role Summary

As a GET (Embedded AI), you'll help take models from dataset → experiment → deployment on Mankomb hardware with tight latency/memory/power budgets. Own the end‑to‑end applied‑AI cycle: collect data, design experiments, write training/inference code, and optimize models to run on‑device (ARM/Linux/RTOS). You'll work hands‑on with sensors, test rigs, and production firmware to make models reliable in real kitchens and field conditions.

What you will do

1) Data Ops & Instrumentation

• Plan and execute data collection on devices (vision/sensors), including checklists, labelling, sampling plans, and consent.
• Build labeling pipelines (e.g., CVAT/Label Studio) and ensure dataset hygiene (splits, versioning, metadata).
• Implement simple data quality gates (blur/occlusion checks, class balance) and maintain a data catalog.

2) Experimentation & Metrics

• Write reproducible experiments (config‑driven) with versioned runs; track metrics (precision/recall/F1, mAP, ROC‑AUC) and confusion matrices.
• Perform hyperparameter search, ablations, and error analysis; summarize results in weekly experiment notes.

3) Model Development

• Implement and train models in Python (PyTorch/TensorFlow) with clean module structure, typing, and tests.
• Apply compression: quantization‑aware training, pruning, distillation. Export to ONNX/TFLite as needed.

4) Deployment on Mankomb Hardware

• Convert/optimize models for edge: TFLite/TFLite‑Micro/ONNX‑Runtime; leverage NEON, CMSIS‑NN, or vendor NPUs where applicable.
• Write/optimize C/C++ inference code, circular buffers, and low‑latency I/O. Profile latency, RAM/Flash, FPS, power, and thermal behavior.
• Integrate with firmware/OS (Linux/RTOS/FreeRTOS), cross‑compile toolchains, and CMake builds.

5) Embedded & Systems Integration

• Implement sensor drivers (I²C/SPI/UART), basic image signal pipeline tweaks, and real‑time scheduling.
• Collaborate with mech/EE teams on fixtures, lighting, shielding, and EMI/EMC‑aware data capture.

6) Quality, Safety & Documentation

• Follow coding standards, unit/integration tests, and static analysis. Maintain runbooks, SOPs, and release notes.
• Contribute to CI pipelines for builds, tests, and artifact packaging.

Must‑have

• Strong in C/C++ and embedded concepts (memory management, interrupts, DMA, RTOS basics).
• Proficient in Python for data/ML (NumPy/Pandas, PyTorch or TensorFlow, OpenCV).
• Comfortable with Linux, cross‑compilers, CMake, Git; basic shell scripting.
• Understanding of model deployment constraints (INT8 quantization, fixed‑point math, on‑device profiling).
• Clear fundamentals in DSA, OOP, numerics, and signals.

Good‑to‑have

• ARM Cortex‑A/M experience; NEON SIMD, CMSIS‑NN; vendor stacks (e.g., NPU/DSP SDKs).
• ONNX, TFLite‑Micro, OpenCL/CUDA (where applicable), or Arm Compute Library.
• Basic MLOps habits: experiment tracking (e.g., W&B), DVC, artifact registries.
• Exposure to computer vision (lighting, lenses, FOV), data augmentation, and synthetic data.

Program structure (6 months)

• Work across Data/ML, Embedded/Inference, Validation/Release.
• Mentorship: Buddy + monthly technical reviews
• Conversion: High performers may receive a full‑time offer.

Working model & benefits

• On‑site role at HSR Layout with device labs; occasional field data collection.
• Tools, dev kits, and mentoring provided; steep learning curve in real‑world edge AI.

Compensation:

• Upto INR 25,000 per month during term as GET
• INR 7,50,000 p.a. CTC subject to confirmation at the end of 6 month term
• Unlimited coffee and conversations

How to apply

Apply on (Apply button) and email with subject: "GET‑AI: Your Name" and include:

1. Resume (PDF)
2. Links to code/repos or experiment reports
3. A short note (3–5 lines) on an embedded/AI project you built, and the latency/accuracy trade‑offs you solved.

Equal Opportunity: We value diversity and are an equal‑opportunity employer. If you love building AI that actually ships on hardware, we'd love to talk.