AI Data Project Manager · Delivery · Toloka AI · Remote US

DiscoverShipAdopt · Experts + Crowd + Tech

Own the client's data solution. Quality, date, cost, a status they can defend.

Toloka's AI Data Project Manager owns client delivery with a virtual technical team. That is Informatica IPS / SEAL — consulting delivery at scale, client in the room, experts and operators on a follow-the-sun clock. Last 24 months I ran the same loop on LLMs I actually measured. Remote US. Available. I do not have a PhD. I have the delivery seat.

Quality, timeline, unit economics, and a status the client can trust. If the weekly is theater, the account is already leaving. I write the path the team can run, then I stay until they run it on Tuesday.

Jeremy Panasuk
Jeremy Panasuk · Sand Creek, WI
7–14XF team I led
19/20Crisis recovered
8h→1hDawn · quality used
$4M+Services in two years
11 yrsClient delivery mgmt
LLM labMeasured agents · RAG

01 — Your posting · my receipts

Client delivery is the job I already did. LLMs are the stack I already measure.

Own delivery

Quality, timeline, cost, transparency

Dawn 8h→1h because it got used. PepsiCo v9→v10 zero data loss. Micron $3M extensions, zero downtime. Status the stakeholder can defend — not a slide they cannot repeat.

Virtual XF team

Engineers, supply, a follow-the-sun clock

Teams of 7–14. Engineers, QA, GCS, offshore, the floor. Named skill-up. Retros were writeups the next morning, not a slide. That is Experts + Crowd + Tech with a weekly the client can trust.

Client in the chair

Consulting is the plus you listed

Clients asked for me by name. "Rock star." Embedded until adoption, then hypercare. Loyalty is Tuesday. Satisfaction is a survey they forget.

02 — How I run a delivery

Discover. Ship. Adopt. Same mill, new stack.

01 · Discover

Decompose before you staff

Environment Summary. Unofficial definitions. Scope, risks, unit of work. I do not wait for a clean spec. I write the path the team can run — the same muscle as a labeling spec the crowd can actually follow.

02 · Ship

Quality gates on a clock

Implementation, throughput, six-class test. QA twin of production. If humans do not run it Tuesday, it did not launch. Unit economics are a number before and after, not a vibe.

03 · Adopt

Leave a team that stays on track

Named skill-up. Process change as a deliverable. TCS kept Mercedes docs. Playbooks are how 19/20 stays 19/20. Hypercare is default, not a phase I skip.

03 — Clock work · two weeks, their metal

Timeline is a promise. I close it on their floor.

Volkswagen Group of America

Complete Grid/HA in two weeks — day early

They wanted the complete stack, not a subset: bonded network / dual NICs, Oracle RAC on InfiniBand, three 8-core PowerCenter servers. Two-week window. I walked the live demo and finished a day early. Four customer consultants shadowed the entire install, taking notes. Parallel Data Quality on the other side of the building was burning down — that was the pressure in the room, not my DQ miss. No dollar on this copy. The close was: they watched it run on their RAC.

Unnamed customer · do not invent a name

Two-week PowerCenter / Data Quality on their metal

Full build on their metal — not a lab VM — plus complete use-case analysis against their data. Distinct from the Volkswagen Grid/HA (named, no dollar). Closed $375k the night before quarter end. The client stays unnamed here. The motion is the point: two weeks, their machines, a number the quarter could count.

04 — Kohl's, not a dollar chip

Capacity, upgrade, AIX off, Senior VPs, then the close.

When I say Kohl's I do not mean a logo on a trusted-by line. That is a data-platform utilization roadmap a client could defend — the same altitude as owning quality, timeline, and unit economics for Toloka.

Roadmap. Built the future data-platform utilization roadmap: Informatica v9 capacity planning and upgrade, including multiple high-availability and Grid components.

Target architecture. Designed the target Grid/HA on commodity hardware (Cisco or IBM) and the cutover off AIX.

Sized to growth. Analyzed current data volumes and projected 3-year growth; sized the platform to that curve.

Briefed through Senior VPs. Delivered the roadmap and presentation through multiple org levels up through Senior VPs.

Close. Closed ~$2 million in Informatica software as a result of that work.

Their data, not the VM. Ran proof-of-concepts when Informatica VM labs were not enough (Data Quality / MDM preview of their own data).

05 — Independent lab · Aug 2024 – present

Measured agents. Honest on the mill I have not run.

4070 · private inference

Same instinct as a data pipeline

TabbyAPI EXL3, Ollama, Open WebUI, Compose + CUDA on RTX 4070 / WSL2. Local Grid Suite: 1.39 → 37.47 tok/s after GPU routing (~27× vs CPU baseline), logged in SQLite. Throughput you can report — not a vibe.

Agents · MCP · RAG

Structured generation with a bound

Systems → architect → validator on local Ollama: structured output, repair retries. Dockroot-MCP so agents see Docker/network, not host-root. n8n RAG: ingest metrics → retrieve history → route (Discord always, WhatsApp on a rank/score gate).

Honest on SFT / eval / red-team: I have not run a labeling marketplace producing thousands of prompts a week. I have owned data quality, throughput, and client delivery on live platforms, and I have spent 24 months on LLM generate / evaluate / bound loops I can show. I switch domains the way IPS switched clients.

06 — Asked for by name

Customer satisfaction is worthless. Customer loyalty is priceless.

I gave that talk at Informatica while still embedded at Dawn Foods in Jackson, Michigan. Satisfaction is the lowest acceptable service. Loyalty is the warehouse still running the one-hour load — and a virtual team that stays on track after I leave the weekly. Read the talk. Bank of America, Dawn, John, the list we held ourselves to. Then Kohl's, Volkswagen two-week Grid/HA, and the unnamed two-week PC/DQ that closed $375k the night before quarter end.

“Jeremy has been a pleasure and an asset to work with. His knowledge not only of the Informatica suite but also at the operating system level has been truly invaluable.”

Dawn Foods · client

“He’s a rock star out at Dawn. The customer loves him and the sales team has been leaning on him hard.”

Informatica IPS · sales

“It is the belief of the infrastructure staff that you have sent one of the best that Informatica has to offer in Jeremy.”

MF Global · infrastructure

“I trusted that Jeremy would deliver, with quality… exceeded the expectations at someone of his grade level.”

Prime Therapeutics
Toloka: own the client's data solution — quality, date, cost, a team that stays on track. Consulting delivery is the job I already did. LLMs are the stack I already measure.