Top Tech Jobs & Startup Jobs in Boston, MA

Optimize and compress large language and vision models for on-device inference. Build distillation and hardware-specific compilation pipelines, benchmark performance across NPU/GPU architectures, and deploy models to edge environments.

Design and integrate vision encoders and audio-native models into agent reasoning loops, optimize streaming latency for voice-to-voice interactions, and architect multimodal RAG systems to retrieve insights from videos and PDFs while ensuring efficient cross-modal alignment.

Design, orchestrate, and optimize large-scale LLM pre-training runs across 1,000+ GPUs using PyTorch/DeepSpeed/Megatron-LM. Improve networking (InfiniBand/RDMA) and memory management, implement 3D parallelism, and automate checkpointing and failure recovery for month-long training jobs on SLURM or Kubernetes GPU clusters.

Design and run adversarial tests on language and multimodal models, build guardrails and real-time filters for autonomous tools, and develop constitutional AI principles and RLHF alignment pipelines to ensure safe AI deployment.

Design and implement high-fidelity spatial mapping and real-time object recognition pipelines. Integrate vision systems with robotic manipulation and navigation, and train and fine-tune Vision-Language-Action models to improve robotic automation.

Lead research developing foundation models for robotics, align text reasoning with low-level control, design zero-shot generalization experiments, integrate models into robotic applications, and publish and present results at conferences.

Lead product strategy and roadmap to convert advanced AI and robotic research (VLA models, agentic RAG) into customer-facing solutions. Define safety, usability, and performance metrics, work with early enterprise adopters for feedback, and coordinate cross-functional teams to deliver complex hardware and deep-tech software products.

Design and implement advanced control algorithms for bipedal locomotion and manipulation. Train RL policies in simulation and deploy them on physical hardware. Perform sim-to-real transfer and analyze gait and kinematics telemetry.

Develop and integrate high-resolution tactile sensing and dexterous manipulation capabilities: process tactile and force-torque data, implement slip-detection and compliant control, train multi-fingered grasping policies, run experiments, and collaborate to deploy solutions on robotic platforms.

Design and implement Visual-Inertial SLAM pipelines, dynamic obstacle avoidance, and predictive path planning. Optimize navigation algorithms for low-compute edge deployment and deploy autonomous navigation on robotic platforms.