AI/ML Engineer

What is Armory?

What we are building requires hard work, tenacity, ownership, creativity, and way more man-hours than an everyday job.

Armory is a new-age defence tech startup building next-generation products to strengthen Indian forces in the modern battlefields.

At Armory, we identify and solve the most critical, complex and specific challenges in modern warfare. By harnessing the power of artificial intelligence, we create innovative defence technologies and products that protect our nation and allies.

Requirements

About this role

You will own the perception and edge AI stack at Armory: data, models and deployment on hardware. The systems you build run in the field, in real scenarios, where reliability is a baseline and a failure has real consequences. This is a builder role. You will train models, write code and ship on real hardware, starting in ambiguity where specs are thin and priorities shift, and turn it into systems that run with clarity and discipline, without losing speed.

Responsibilities

• Build and train computer vision and multimodal models for detection, tracking and segmentation in real defence scenarios.
• Own the data pipeline: collection, labelling, synthetic data and failure-driven retraining.
• Optimise and deploy models on edge hardware (Jetson, ARM) within strict latency and power budgets.
• Integrate perception with hardware, RF and sensor systems so the whole stack works in the field, not just in the lab.
• Work with hardware, systems and design teams to ship new capabilities, and mentor junior engineers.

Must haves

• 4 to 7 years building and shipping ML systems, with 3+ years in computer vision or robotics perception.
• Strong Python and PyTorch, with working C++.
• Hands-on with detection and segmentation models (YOLO, RF-DETR, SAM) and fine-tuning them on custom data.
• Edge deployment experience on Jetson or similar: TensorRT, ONNX, quantisation.
• Ownership: you take an ambiguous problem to a working system and explain your trade-offs simply, without ego.

Bonus points for

• Sensor fusion, stereo vision, LiDAR point clouds or SLAM.
• Experience with VLMs or VLAs, ROS and simulation (Isaac Sim, AirSim).
• Drone, UAV or geospatial imagery work.
• Open-source contributions, publications or patents, and a keen interest in defence.

What will make you successful in this role?

• Deep technical expertise, worn lightly. You are exceptional without being cocky.
• Shipping model improvements to production every week.
• Bringing order to chaos without adding bureaucracy.
• Getting models running on real hardware in record time.

How will we set you up for success in this role?

• A live project: You will jump straight into a live project as soon as you join. No dummy or test projects.
• Access to resources: You name it, we will do everything to get it for you. That includes GPUs, data and hardware.
• Supportive team: You will be surrounded by the best minds ready to help you.
• Founder access: You will work directly with the founder on what to build and how to build it.

Hiring Process

We target to close the hiring process within 14 days of your application.

• Round 1: Screening Application and HR Round (30 minutes)
• Round 2: Culture Connect (30 minutes)
• Round 3: Founder Round (1 hour)
• Round 4: In-Person Meet in Gurgaon (1 hour)

Work Location

Armory highly values having employees working in the office. We are based out of Gurgaon and require employees to be in the office.

Benefits

What we expect from you

• Genuine curiosity
• Founder's mindset
• First principles thinking
• Execution, not just ideas
• Bias for action
• High energy
• High agency
• Highest level of integrity
• Getting your hands dirty
• Audacious state of mind
• Humility

What to expect from us

• Team of A players
• Work directly with the founder
• Ridiculous levels of responsibility
• Fast and decisive approach
• Work on hard problems

What your first 3 months at Armory will look like

30 days

• Complete onboarding, learn our mission and values, meet your team, and set up your development environment.
• Audit the current models, datasets and inference stack.
• Contribute to a live project and ship a model improvement to production.

60 days

• Deploy an optimised model on target edge hardware, with benchmarked latency, accuracy and power numbers.
• Fix what is fragile: set up dataset versioning, evaluation benchmarks and a repeatable train-to-deploy pipeline.
• Showcase a working prototype of a new perception capability.

90 days

• Own a major perception module in production, end to end: data, training, evaluation and field deployment.
• Define the field-validation test plan and optimise performance based on the results.
• Present your roadmap for the AI stack and discuss your growth plan with the founder.