Digantara is a leading Space Surveillance and Intelligence company focused on ensuring orbital safety and sustainability. With expertise in space-based detection, tracking, identification, and monitoring, Digantara provides comprehensive domain awareness across regimes, allowing end users to have actionable intelligence on a single platform. At the core of its infrastructure lies a sophisticated integration of hardware and software capabilities aligned with the key principles of situational awareness: perception (data collection), comprehension (data processing), and prediction (analytics). This holistic approach empowers Digantara to monitor all Resident Space Objects (RSOs) in orbit, fostering comprehensive domain awareness.
Digantara is seeking an Embedded software, AI & Vision algorithms Intern to benchmark and characterise image processing and machine learning algorithms on embedded platforms with multi-core CPUs, GPUs and NPUs running Linux. The intern will help determine how on-board processing algorithms should be mapped to heterogeneous compute engines to achieve the best performance, power efficiency and accuracy.
Why Us?Competitive incentives, galvanizing workspace, blazing team, frequent outings—pretty much everything you have heard about a startup + you get to work on space technology.
Hustle in a well-funded startup allowing you to take charge of your responsibilities and create your moonshot.
Ideal Candidate:
A curious, hands-on candidate with a strong grounding in Linux architecture, keen to understand how vision and ML algorithms behave on embedded hardware.
ResponsibilitiesBring up and configure embedded Linux boards, including drivers, runtimes and SDKs, for benchmarking.
Design benchmark plans for image processing pipelines (resize, colour conversion, filtering, ISP-style stages, optical flow) and ML models (classification, detection, segmentation).
Map workloads across CPU cores, GPU and NPU/DSP, and compare execution on each, including heterogeneous split pipelines.
Measure end-to-end and per-stage latency, throughput (FPS), jitter, CPU/GPU/NPU utilisation, memory bandwidth and footprint, power and thermal behaviour.
Use Linux profiling tools to locate bottlenecks.
Evaluate the effect of quantisation (FP32/FP16/INT8), batch size, input resolution, threading, CPU affinity, DVFS governors and zero-copy buffer sharing.
Validate accuracy against reference results so that performance gains do not hide quality loss.
Automate repeatable benchmark runs, keep results traceable to hardware and software configuration, and document findings with clear recommendations.
Pursuing B.Tech/B.E or M.Tech/M.E/M.S in Electronics, Electrical, Computer Science or a related discipline.
Solid grasp of computer architecture: caches, memory hierarchy, SIMD, multi-core execution, and how GPUs and NPUs differ from CPUs.
Working knowledge of Linux: shell, processes and threads, file systems, building and cross-compiling software.
Programming in C/C++ or Python sufficient to run, modify and script experiments.
Basics of digital image processing (convolution, colour spaces, filtering, scaling) and neural networks (CNN layers, inference vs training).
Understanding of measurement fundamentals: averages vs percentiles, warm-up, variance and controlled experiments.
Experience with vision libraries such as OpenCV, GStreamer, V4L2 or OpenVX.
Exposure to inference runtimes such as TensorRT, TensorFlow Lite, ONNX Runtime, OpenVINO, TVM or vendor NPU SDKs (TIDL, eIQ, SNPE/QNN, RKNN).
Knowledge of parallel programming with CUDA, OpenCL, Vulkan compute, NEON intrinsics, OpenMP or pthreads.
Familiarity with PyTorch or TensorFlow, ONNX model export and INT8 quantisation.
Understanding of Linux internals: device tree, kernel modules, schedulers, cgroups, DVFS/cpufreq, DMA-BUF.
Prior projects on Raspberry Pi, Jetson or similar boards, and use of Yocto/Buildroot, Docker, Git and Python data tools.
Ability to work in a mission-focused, operational environment.
Ability to think critically and make independent decisions.
Interpersonal skills to enable working in a diverse and dynamic team.
Maintain a regular and predictable work schedule.
Writing and delivering technical reports and briefings.
Verbal and written communication skills as well as organisational skills.
Duration: 6 months (provisionally extendable), full time. Job Location: Hebbal, Bengaluru (on-site, lab-based). Work Type: WFO with alternate Saturdays.
Skills Required
- Currently pursuing a B.Tech, B.E., M.Tech, M.E., or M.S. in Electronics, Electrical Engineering, Computer Science, or a related discipline
- Understanding of computer architecture, including caches, memory hierarchy, SIMD, multicore execution, and CPU/GPU/NPU differences
- Working knowledge of Linux, including shell, processes, threads, filesystems, software builds, and cross-compilation
- Programming experience in C, C++, or Python sufficient to run, modify, and script experiments
- Fundamental knowledge of digital image processing and neural networks
- Understanding of measurement fundamentals, including averages, percentiles, warm-up, variance, and controlled experiments
- Experience with OpenCV, GStreamer, V4L2, or OpenVX
- Exposure to inference runtimes or vendor NPU SDKs, including TensorRT, TensorFlow Lite, ONNX Runtime, OpenVINO, TVM, TIDL, eIQ, SNPE/QNN, or RKNN
- Knowledge of parallel programming with CUDA, OpenCL, Vulkan compute, NEON intrinsics, OpenMP, or pthreads
- Familiarity with PyTorch, TensorFlow, ONNX model export, and INT8 quantization
- Understanding of Linux internals, including device trees, kernel modules, schedulers, cgroups, DVFS/cpufreq, and DMA-BUF
- Prior projects using Raspberry Pi, Jetson, or similar embedded boards
- Experience with Yocto, Buildroot, Docker, Git, and Python data tools
- Ability to work in a mission-focused operational environment and make independent decisions
- Technical report and briefing delivery skills
- Verbal, written, interpersonal, and organizational skills
- Ability to maintain a regular and predictable work schedule
What We Do
Digantara is an Indian company specializing in space situational awareness, intelligence, and surveillance. It develops infrastructure and data solutions for monitoring objects and traffic in Earth orbit, and is expanding its capabilities into missile defense.
