Research Scientist — Neuromorphic Computing & Efficient AI

Research Scientist — Neuromorphic Computing & Efficient AI
Location: Bengaluru, India Type: Full-time Team: Autonomous Machines Applied Research Center Eligibility: PhD, or MTech with relevant research/industry experience
About the role
We’re building efficient AI systems for autonomous machines — spanning spiking neural networks, event-based vision, quantized/compressed model architectures, and reinforcement learning — with the eventual aim of getting research results onto real, resource-constrained hardware (FPGAs and neuromorphic accelerators). You’ll work on problems that span the end to end pipeline - algorithm design, training methodology, and hardware deployment.
This is a research role: you’ll be expected to contribute to publications and Intellectual property and also see ideas through to working demos and systems on real silicon. We’re open to both PhD holders and MTech candidates who’ve built up equivalent depth through hands-on research or industry experience — what matters most is demonstrated ability to do original, rigorous technical work in this space. The candidate should ideally have prior experience with publishing research and/or developing Intellectual property. Being able to work with hardware (FPGAs, Neuromorphic chips/Raspberry pi/NVIDA jetson platforms) would be a huge plus.
What you’ll work on
Spiking neural networks — architecture design, training methodology (surrogate gradients, quantization-aware training), and both dense and sparse/event-driven formulations for video and sensor data
Event-based and neuromorphic vision — pseudo-event and true event-camera pipelines
Model compression and quantization — ternary/binary weight networks, hardware-aware, quantization-aware training, for real-time inference under tight power and area budgets
FPGA and neuromorphic hardware deployment — taking trained models to hardware (Xilinx KV260 / Zynq UltraScale+, and neuromorphic accelerators such as BrainChip Akida and SynSense Speck), including custom HLS/RTL for architectures that don’t map cleanly onto existing toolchains
Reinforcement learning and control — active inference / free-energy-minimization approaches to RL, applied to robotic control (manipulator reaching, drone hover control) and classic control benchmarks
Applied computer vision for autonomous systems — action/intent recognition, detection, and classification pipelines for real-world deployment scenarios
Who we’re looking for
Required — one of the following: - PhD in Computer Science, Electrical/Electronics Engineering, Computational Neuroscience, or a closely related STEM field. PhDs with quantitative research experience in any field involving Machine Learning/Artificial Intelligence methods are encouraged to apply. or - MTech in a related field with at least 4 years of hands-on research or industry experience in one or more of the areas below, demonstrated through shipped work, publications, open-source contributions, or a strong project portfolio.
Demonstrated depth in one or more of: Efficient/compressed deep learning, reinforcement learning, or computer vision — via publications, patents, or substantial hands-on project work - Solid hands-on experience with PyTorch, including building custom training pipelines and non-standard model architectures from scratch - reading and reasoning about someone else’s research codebase, with experience in implementing SOTA methods from github repositories. The candidate must be a quick learner and be able to quickly adopt and implement AI methods based on varied project requirements.
Preferred (any of the following are a strong plus, not all required): - Experience with autonomous systems (ADAS systems/Drones etc.) - Experience with SNN training frameworks (SpikingJelly, BrainCog, snnTorch, or similar) - FPGA development experience — Vitis/Vivado HLS or hls4ml-style ML-to-FPGA toolchains - Experience with neuromorphic hardware (BrainChip Akida, SynSense Speck, Intel Loihi, or similar) – Experience with edge computing solutions like the NVIDIA Jetson platform - Background in computational neuroscience - Experience with model quantization (ternary/binary weight networks, QAT) for edge deployment - Robotics experience — ROS, manipulator or drone control, sim-to-real pipelines (e.g. CoppeliaSim, Gazebo)
Show more Show less