Principle AI Engineer

Job Description Principal AI Engineer Magnasoft · Bengaluru · Hybrid ​ Hands-on technical leadership · Scaling our AI/ML engineering org Why this role exists Magnasoft is twenty years into building one of the world’s deepest geospatial data assets — and is now turning that asset into AI-powered software products. We have the data, the domain, and the customer relationships, and our first product is live in beta with early deployments validated and commercial traction building. What we’re building next is the engineering organization that turns this from a services company into an AI-product company. Today, that AI/ML engineering work runs through a single person: our VP & Head of Technology. This role exists to change that. As Principal AI Engineer , you become the most senior engineer in the AI org and the technical owner of how we build — the architecture, the delivery, and the engineering craft of the team. This is a player-coach seat: you set the technical direction and you stay in the code. You are not inheriting a mature platform to maintain — you are scaling it from a strong foundation. If you want to be one slice of a large, settled ML org, this isn’t it. If you’ve drifted away from the keyboard and want a pure management seat, this isn’t it either. If you want to define — and personally build — the technical spine of a company at its inflection point, read on. What you’ll own Architecture & technical direction. You own the architecture of the ML systems behind our product portfolio — a shared computer-vision / OCR / human-in-the-loop engine applied across multiple product domains. A key architectural mandate: design the path from periodically-retrained static models toward models that improve continuously from a production feedback signal , with the right safety gates rather than a manual blessing on every release. Hands-on engineering. You write production code — and not only model code. You build the product code the models live in : the AI backend that runs the models, the post-processing and pipeline logic, and the services and data layer around them (all in Python). You’re as at home in the pipeline, the data layer, and the APIs as in the modeling code. You set the bar by example and multiply the team’s output with modern AI-assisted / agentic coding workflows (e.g., Claude Code). The team should get faster because you’re on it — through your architecture and how you actually build. Technical delivery across the portfolio. You own the engineering bar and the delivery of the AI/ML work across our active product lines. You make the build-vs-buy calls, set standards for how models ship, and are accountable for the systems being reliable enough to charge customers for. The data flywheel. Our advantage is a deep proprietary data asset and a feedback loop that turns production into training signal. You’ll partner closely with the Senior Applied ML Engineer (who owns model-quality methodology and the eval harness) to make that loop real: data and label versioning, a trustworthy automated evaluation gate, and the retraining cadence that lets models improve release over release. Technical leadership & mentorship. You are the technical authority the AI/ML engineers work under day to day, across the whole stack — models, the AI backend, the pipelines, and the data layer — architecture, design, code review, and mentoring. You raise the team’s ceiling and make it able to run without a single point of failure. People-management (performance, 1:1s, career) sits with the VP & Head of Technology, so your energy goes to the engineering, not the org chart. Building the AI team. You help build the team you’ll lead. As we scale the AI/ML org, you own the technical side of hiring — designing how we interview AI/ML engineers, running technical interviews, setting the bar for who clears it, and growing a senior technical bench so the team is never one person deep. The final hire decision and offers sit with the VP & Head of Technology; getting the technical judgment right is yours. Near-term, ramping down: technical stabilization of an existing delivery program. For the first phase you’ll own the technical stabilization and clean handover of an existing delivery program to a maintenance footing. A focused, time-boxed responsibility (~10–25% early on) that decays toward near-zero as it stabilizes; the operational coordination sits with delivery/operations. This is a sunset duty, not a standing one — your gravity is firmly on the forward-looking product and platform. Requirements What we’re looking for (must-haves) • ~10–12 years building and shipping software overall, with ~6–8 years hands-on in production AI/ML , including technical leadership of engineers as a lead, principal, or architect who owned a team’s technical direction and stayed technical . We index on the depth and recency of hands-on AI/ML ownership, not raw tenure. • Hard-core hands-on. You still write code daily and want to. This is not a role for someone who has fully transitioned to management. • Deep applied ML expertise — ideally in computer vision, OCR, or document/spatial AI . You’ve taken models from prototype to production at scale, with genuine evaluation discipline (you talk about model quality in terms of cost-of-error and unit economics, not just accuracy). • Experience architecting and operating continuous or online retraining loops in production — models that improve from a live feedback signal — or a clear, demonstrated path toward them from batch/periodic retraining. • Strong systems and architecture chops on a major cloud, AWS preferred — solid command of commonly used AWS services and design patterns, scalable Kubernetes deployment (we run AWS EKS ), and GPU-based architecture for AI workloads (training and inference at scale). • Full-stack AI-backend architecture — a level above the engineers you’ll lead. You own not just the product code around the models (invocation, post-processing, pipeline stages) but its architecture — the pipeline design and the data model — and you review the team’s schema, aggregation, and pipeline work with authority. You can’t set a bar you’re standing below, so on this stack you’re ahead of the engineers, not level with them. Our stack includes a document store (MongoDB) and PostgreSQL , containerized with Docker on AWS EKS . Deep command of a document store at design level is strongly expected; MongoDB specifically is not a hard requirement — equivalent NoSQL plus the ability to own ours quickly is fine. • A track record of being the senior technical authority people escalate to — and of building practices, not just code. Our stack Python across the board — modeling and the AI backend / pipelines; PyTorch for modeling; a document store (MongoDB) and PostgreSQL ; Docker / Kubernetes on AWS EKS ; AWS for cloud and GPU-backed training/inference; React on the front end. You don’t need every box ticked — depth in the ML and the Python backend/data layer matters most, and we expect the right person to pick up adjacent tools quickly. Strong plus (any of these moves you up the stack) • Fluency with AI-assisted / agentic coding methodologies (e.g., Claude Code) and a track record of using them to materially increase engineering throughput — yours and the team’s. • Geospatial / GIS experience — GDAL/geopandas, remote sensing, point-cloud (PDAL, Open3D). A real advantage, but we’ll happily ramp the right person on the domain. • Agentic systems / orchestration , or experience with autonomous model-update or self-improving loops. • Production human-in-the-loop systems — where human corrections feed the next model version. • Experience productizing services into software (services-as-software, platformization). • Self-supervised / foundation-model work on proprietary datasets . You might not be a fit if • You have no interest in moving beyond batch retraining toward live, continuous systems — that direction is central to the role. • You’ve moved away from hands-on coding and want a pure people-management seat. • You need a mature ML platform already in place. Here, you build it. • You’re deep in modeling but thin on backend and data-layer engineering — this role architects and reviews the product code, pipelines, and data layer, so model expertise alone isn’t enough. • You’re looking for a research-and-publish role. Team & reporting • Reports to the VP & Head of Technology . • You hold functional technical authority over the AI/ML engineers — architecture, design, code review, mentoring — and partner closely with the broader engineering org (Backend, Frontend, Full-stack, DevOps, QA). • This is the senior-most technical seat in the AI/ML org: you shape how the team builds, not just what it builds. Location & work mode • Bengaluru-based. Hybrid — up to ~40% work-from-home (roughly 3 days/week in office). How to apply Interested, or know someone who’d be a fit? Apply through this posting or reach us at [email protected].