Job description

About NegotiateAI (Menlo Ventures Backed) Today we're building AI for category managers at manufacturing and industrial companies, taking the grind out of indirect spend and delivering enterprises millions in savings. That's the starting point. Where we're going is general procurement AI that spans the entire function. About this role Procurement data is a mess, and every enterprise's mess is different. Part numbers don't match, supplier names are spelled four ways, and the ERP is a decade of customizations on top of SAP. Our platform has to work inside that reality on day one, not after a six month cleanup project. You'll be the first Forward Deployed Engineer at NegotiateAI: embedded with our earliest customers, building until the product delivers measurable savings against their real data, then pushing what you learned back into the core platform. This is an engineering role, not a solutions or consulting role. You write production code and you own what you ship. What you'll do - Embed with customer procurement and IT teams to map how spend actually flows: intake, approvals, category taxonomies, supplier master data, and where the savings are hiding. - Define crisp success metrics with the customer before you build. Realized savings, categorization accuracy, cycle time, coverage. Then instrument them so nobody has to argue about whether it worked. - Build integrations into customer systems: SAP, Ariba, Coupa, Oracle, NetSuite, data warehouses, SSO and identity, and the inevitable SFTP drop. Navigate read-only versus write-back postures and design for whichever one the customer will actually approve. - Go from a raw spend extract to a working pipeline fast: entity resolution on suppliers, part identifier normalization, category mapping, and ground truth you can defend in a room full of skeptics. - Model customer spend in ClickHouse, which backs our analytics and benchmarking layer, and keep queries fast as datasets grow from one customer's category to an enterprise's full indirect spend. - Customize the agentic workflows at the core of the product for each deployment: retrieval sources, tools, prompts, and guardrails, then harden them for reliability and observability. - Build per-customer evals against their own data so we can prove accuracy, catch regressions, and show progress week over week. - Drive adoption. Run training, write runbooks, and hand off something durable to a champion inside the account instead of becoming the single point of failure. - Be the technical voice in front of prospects: lead demos on sales calls, scope pilots against a prospect's own data, and handle security reviews, data handling questions, and SOC 2 questionnaires. - Feed patterns back into the platform. Recurring integrations, failure modes, and workflows should become product, not a folder of one-off scripts. Must haves - 4+ years of professional software engineering experience with a track record of shipping production systems end to end. - Real comfort working directly with customers, from scoping an ambiguous problem with a category manager to running a training session to answering a hard question from a chief procurement officer or CIO. You treat time with stakeholders as part of the engineering work rather than a detour from it. - Hands-on experience building with LLMs in production: agentic workflows, tool use, retrieval, structured outputs, and evaluation. - Strong backend and integration skills. API design, data pipelines, and working against systems whose documentation is wrong. Depth matters more to us than experience in any particular language. - Experience shipping against messy real-world data: cleaning, normalization, entity resolution, and establishing ground truth. - Judgment about what stays bespoke and what becomes product. Custom work is how we learn, but a deployment that only works for one customer is a liability. - Comfort with early-stage ambiguity: a small team, shifting priorities, and shipping before everything is fully specified. - Willingness to travel to customer sites, primarily manufacturing and industrial companies, roughly once or twice a month depending on the deployment. Nice to haves - Experience integrating with enterprise procurement or ERP systems such as SAP, Ariba, Coupa, Oracle, or NetSuite. - Background in procurement, supply chain, sourcing, or manufacturing software. - Deep production experience with TypeScript. - Experience designing agent harnesses: the scaffolding that makes a model reliable in production. - Experience with ClickHouse, or with other columnar analytics stores such as BigQuery, Snowflake, or Databricks. - Experience meeting SOC 2, GDPR, or similar compliance requirements at a startup. - Prior forward deployed, deployment strategist, solutions engineering, or technical consulting experience at a company like Palantir, Scale, or a comparable enterprise AI team. - Experience at an early-stage startup (seed through Series B). How you will make an outsized impact - Customer obsession over elegance. You spend as much time understanding why a category manager does something the hard way as you do building the fix. The best architecture that nobody uses is worth nothing. - Ship, then learn. You get something in front of a real user quickly and let their reaction shape the next version. Progress is measured in what got deployed, not what got planned. - Own the outcome. Your responsibility doesn't end at the code you shipped. If a deployment is stalled on a data access request or a procurement team that hasn't been trained yet, unblocking it is part of the job. - Generalize deliberately. You know when to hardcode something to unblock a customer this week, and you know when to stop and build the abstraction. You leave the platform better than you found it.

Requirements

Must-haves

About NegotiateAI (Menlo Ventures Backed)
Today we're building AI for category managers at manufacturing and industrial companies, taking the grind out of indirect spend and delivering enterprises millions in savings.
That's the starting point. Where we're going is general procurement AI that spans the entire function.
Embed with customer procurement and IT teams to map how spend actually flows: intake, approvals, category taxonomies, supplier master data, and where the savings are hiding.
Define crisp success metrics with the customer before you build. Realized savings, categorization accuracy, cycle time, coverage. Then instrument them so nobody has to argue about whether it worked.
Go from a raw spend extract to a working pipeline fast: entity resolution on suppliers, part identifier normalization, category mapping, and ground truth you can defend in a room full of skeptics.
Model customer spend in ClickHouse, which backs our analytics and benchmarking layer, and keep queries fast as datasets grow from one customer's category to an enterprise's full indirect spend.
Customize the agentic workflows at the core of the product for each deployment: retrieval sources, tools, prompts, and guardrails, then harden them for reliability and observability.
Build per-customer evals against their own data so we can prove accuracy, catch regressions, and show progress week over week.
Drive adoption. Run training, write runbooks, and hand off something durable to a champion inside the account instead of becoming the single point of failure.
Be the technical voice in front of prospects: lead demos on sales calls, scope pilots against a prospect's own data, and handle security reviews, data handling questions, and SOC 2 questionnaires.
Feed patterns back into the platform. Recurring integrations, failure modes, and workflows should become product, not a folder of one-off scripts.

Nice-to-haves

Experience integrating with enterprise procurement or ERP systems such as SAP, Ariba, Coupa, Oracle, or NetSuite.
Background in procurement, supply chain, sourcing, or manufacturing software.
Deep production experience with TypeScript.
Experience designing agent harnesses: the scaffolding that makes a model reliable in production.
Experience with ClickHouse, or with other columnar analytics stores such as BigQuery, Snowflake, or Databricks.
Experience meeting SOC 2, GDPR, or similar compliance requirements at a startup.
Prior forward deployed, deployment strategist, solutions engineering, or technical consulting experience at a company like Palantir, Scale, or a comparable enterprise AI team.
Experience at an early-stage startup (seed through Series B).
How you will make an outsized impact
Customer obsession over elegance. You spend as much time understanding why a category manager does something the hard way as you do building the fix. The best architecture that nobody uses is worth nothing.
Ship, then learn. You get something in front of a real user quickly and let their reaction shape the next version. Progress is measured in what got deployed, not what got planned.
Own the outcome. Your responsibility doesn't end at the code you shipped. If a deployment is stalled on a data access request or a procurement team that hasn't been trained yet, unblocking it is part of the job.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aigoapitypescript

Platform: ashby
Effort: quick