Metadome is hiring a Technical AI Solutions Engineer to own the technical side of enterprise AI pursuits and implementation programs across our OEM customers, and to turn them into scoped, priced, and delivered implementations with the sales team.

Metadome builds CAD-to-digital-twin and AI platforms for automotive, heavy-equipment, and industrial OEMs. Our technology powers sales and aftersales for 30+ automotive and heavy-equipment OEMs across the US, India, and the Middle East. Our platform ingests CAD, PLM metadata, parts catalogs, and service manuals and produces digital twins, configurators, service content, and vision models on top of them.

We expect an engineer in the room who can architect the solution, size the compute, run the pilot, and answer the security and data-ownership questions on the spot.

The role sits between sales and engineering. You are the technical owner of several pursuits at once, from first call to pilot sign-off: you shape the solution with the customer's engineers and consulting partners, define what is built, run the pilot to a kill-or-close decision, and hand a clean build to the delivery team. Pilots without a commercial track die; you run both tracks together.

You own the technical side of multiple enterprise AI pursuits at once, from discovery to pilot close, across automotive, heavy-equipment, and industrial customers in India and the GCC.

Pre-sales solutioning

Run technical discovery with customer marketing, digital, engineering, IT, and dealer teams; map their data (CAD, PLM, parts catalogs, manuals, job cards, images) to what our platform can do today and what needs building.
Design the solution architecture for whatever the problem needs, from computer vision and LLM or RAG assistants to generative content pipelines, digital-twin integration and agentic workflows: model choice (open-source vs hosted), context layer, data pipeline, edge vs cloud split, GPU sizing and concurrency for stated volumes.
Build and deliver live demos on customer data within days, not weeks; own the demo environment and sample datasets.
Answer the security, data-ownership, and IP questions in the room: hosting model, subprocessors, SOC 2 Type II / ISO 27001 evidence, what stays with the customer and what stays with Metadome.

Pilot and program leadership

Write the SOW: scope, success metrics, data the customer must supply, timeline, kill-or-close gate.
Run the day-to-day: weekly cadence with the customer, data evaluation, model training and eval loops, benchmark reports, and the closing readout to the business sponsor.
Work inside the customer's program structure, including their system integrators and consulting partners, without letting them own the architecture of what we build.
Hand over a documented build to delivery and product once the pilot converts.

Commercial hygiene

Draft the technical sections of proposals and RFQ responses; price to pricing.metadome.ai as the canonical source, framed as plans and development scope, never per-part rates.
Keep a live view of technical risk and dependency per deal for the sales forecast; flag pilots that are drifting without a commercial track.

Repeatability

Turn each pursuit into reusable assets: reference architectures, demo kits, a packaged security and data-ownership answer, pilot templates and eval harnesses.
Feed field learnings back to product and the CAD AI team as prioritised gaps with evidence, not opinions.

The portfolio is broad by design: you will move between these families in the same week, for different customers at different stages, and are expected to be credible in all of them within a quarter.

Configurators and marketing content:

Photorealistic, CAD-accurate configurators; video-first product discovery; personalised journeys; measurable engagement and conversion uplifts with AI

CAD intelligence and digital twins:

Agentic digital production

Context and retrieval layers, model portability, guardrails, cost per query

Platform integration and deployment

Hosting on the customer's cloud or GPUs; integration with DMS, CRM, PLM and eCommerce; security and data-ownership terms

Deployment architecture, SOC 2 / ISO 27001 evidence, IP and hosting positions that sales can sign

By day 180 you carry several pursuits at once, have converted at least one pilot into a priced program, and have left behind a repeatable pursuit kit.

Day 30

Take over the live pursuits from the founders, each with a written scope, success metrics, data asks, a compute estimate, and a kill-or-close date. Run at least one live demo on a customer's own data.

Day 90

Two or more pilots running on a weekly cadence across different solution families, each with an eval report against an agreed baseline and a commercial proposal on the table alongside it. Packaged security and data-ownership answer in use across accounts.

Day 180

At least one pilot converted into a priced program, and every pilot closed with a documented decision (convert, extend or kill); a pursuit template (scope, eval harness, proposal skeleton) used on every new deal; a reference architecture and demo kit per solution family in the repo, used by sales without you in the room.

You are measured on pilot-to-contract conversion, the number of pursuits you carry concurrently, time from first call to live demo, forecast accuracy on the technical risk you flag, and the quality of hand-offs to delivery.

Experience:

6–10 years in applied ML, computer vision or AI engineering; 2+ years in a solutions, forward-deployed or technical pre-sales role; has run several AI implementation programs, of more than one kind, end-to-end with enterprise customers, from scope and data through build, evaluation and production

Startup experience selling into Indian or US OEMs; time at a GPU cloud, model provider, or CV product company

Vision and multimodal models

Fine-tuning (LoRA and full), evaluation design (IoU, mAP, calibration), synthetic-data generation, and the trade-offs between hosted APIs and self-hosted open-source models

Shipped defect or damage detection from field images; image-to-3D or novel-view synthesis; NVIDIA Nemotron/Omniverse-class tooling

LLM and agent systems

Retrieval and context layers over enterprise documents and records; tool use and agent orchestration; evaluation harnesses for generative output; cost and latency control per query

Multi-agent pipelines in production; guardrails and red-teaming; fine-tuned small models for domain tasks

Inference and infrastructure

Has deployed inference at production scale and can size it live: GPU selection, concurrency, latency budgets, edge versus cloud, cost per request

Customer-hosted or on-prem deployments; MLOps that runs training and eval loops unattended

Enterprise data

Builds over the data enterprises actually have: CAD and PLM metadata, parts catalogs, PDFs and scanned manuals, image sets with inconsistent labels, CRM and DMS records; knows what a context or retrieval layer costs to build and maintain

STEP/JT/FBX; Siemens Teamcenter or similar PLM; Blender or Unreal pipelines

Engineering

Python first; production-grade code, containers, GCP or AWS

Agentic tooling: MCP servers, orchestration frameworks, agent evaluation

Solutioning and pilots

Has written pilot or POC scopes with success criteria and taken them to a commercial decision; can explain why a pilot died and what they would change

Has worked alongside large system integrators (Infosys, Deloitte, Accenture) on a customer program without ceding the technical design

Security and commercial

Fluent in SOC 2 / ISO 27001 controls, data residency, subprocessors, and model and data ownership; can write it for a procurement team; co-owns proposals with sales

RFQ and RFP responses to Indian OEM

Domain:

Comfortable across automotive, heavy-equipment, and industrial OEM environments, on both the sales and marketing side and the service and after-sales side; dealer operations: job cards, labour codes, warranty, parts identification, DMS platforms.

Location

Bengaluru or Delhi NCR; travel to customer sites in India and occasionally the US

Willing to relocate to Bengaluru

Breadth matters more than depth in any single model: in a typical week you will move between a vision pilot, a generative content pipeline and an enterprise assistant, for different customers, at different stages, in different industries.

How the role works

You are paired with a sales owner on every pursuit and draw on the CAD AI and platform teams for build capacity; you do not carry a quota, but your pilots do.

Reports to: COO, with a dotted line to the CEO for pursuit priorities and forecast.
Works daily with: the COO and the Strategic Account Director for India and GCC, the Head of Product (DP platform), the configurator PM, and the CAD AI team for model work.
Customer-side counterparts: digital and martech leads, dealer-operations teams, IT security and procurement, and the customer's consulting partners.
Cadence: weekly pipeline review with sales, weekly program review per live pilot, monthly readout to the CEO on pursuit health and conversion.
Tools: Python, GCP (with Azure/AWS on customer request), Sprinto for compliance evidence, HubSpot for pipeline, Slack, Fireflies for call notes, pricing.metadome.ai for commercials, Claude Code for build velocity. We work AI-native; you are expected to automate the repetitive parts of the job.
Location and travel: Bengaluru HQ preferred, Delhi NCR acceptable given the OEM customer base there; expect 30–40% travel to customer sites in India

AI Solutions Architect- Enterprise