Agentic cost engineering platform for manufacturing.

Apex reads the engineering drawing, works out what the part should actually cost from live cost-driver data, and guides the cheapest, lowest-risk way to make or source it.

What we're building

Product Mockup

0:00 / 0:00

Rotate your phone for the best view

What you are watching: a buyer types Cost this horn assembly. Target volume 50,000 a year, sourced from Gurgaon, landed to Detroit. Apex reads the drawing, maps all 28 parts inside it and returns the should-cost, with every line visible.

should-cost $1.58 confidence ±4% at 95% 28 child parts 9 process steps, each costed

Cost stack

Child parts, process, overhead, profit, outbound. Every line costed, every equation open.

Sensitivity

The five drivers that move this number: copper wire rate, labour, volume, power, tariff.

Suggestions

Cheaper material and process moves, each backed by peer evidence and live rates.

Fact pack

The negotiation document, ready before the supplier walks in.

01 · The problem

The number is real but in the domain of a few experts today.

Every part has a true cost of making. Manufacturers call it the should-cost, and it decides who wins the negotiation.

Material Machine time Labour Logistics Energy Margin

The supplier

Knows this number cold. It is their business.

So most sourcing decisions are negotiated with one side blind, and it is rarely the supplier.

Providing a real time, accurate and defensible should-cost number in the hands of designers and buyers dramatically shifts the balance of the negotiation back to the buyer.

This is what Apex is building.

02 · What we're building

Cost intelligence, not an AI wrapper.

The question is not "how do we automate the estimate". It is "how do we make the estimate accurate, defensible, and continuously current".

Apex Data

Layer 1 · The bedrock.

A live, city-by-city map of what making things costs. It never stands still.

Machine rates, wages, materials, freight, energy and tariffs.
Primary sources only, never averaged, cleaned into golden records that trace back to source.
Your own ERP records and past quotes train your instance alone: ring-fenced, confidential, harder to leave.

The data competitors cannot fake.

Apex Cost

Layer 2 · Hero product.

Reads the drawing, lists every part inside it, builds the should-cost bottom-up.

Globally proven, transparent engineering equations, not AI guesswork. Every equation visible.
Days of expert work take minutes.
Tune the models to your own plants in the interface, no software ticket. Electronics modelled natively.

Deterministic, so it is defensible.

Apex Suggest

Layer 3 · The killer feature.

Learns from every estimate and sourcing decision it has ever seen.

Surfaces the cheaper material, process, supplier or location before you ask.
Evidence attached: comparable parts, past decisions, anonymized market patterns, savings shown.
Coaches engineers inside their design tools, before an expensive choice gets locked in.

Fully traceable, never a black box.

Apex Data → Apex Cost → Apex Suggest → back into Apex Data

Three layers, one system. Each layer makes the other two better.

Built for the three teams who decide what a part costs.

Cost engineers Sourcing Procurement

Today, a formal should-cost needs a scarce specialist. Apex puts the same quality of analysis in every buyer's hands, turning a specialist bottleneck into a generalist skill. What makes that possible now is timing.

03 · Why now

This was not buildable two years ago.

Three shifts converged. Each one removes a reason this has failed before.

Agents can finally do the work.

Agentic AI now automates the full cost-engineering and procurement workflow, end to end. Complex work that took expert teams days simply could not be automated before.

Supply chains are being redrawn.

Manufacturers are rearranging where parts are made and sourced across the globe. Knowing the true cost in every region has become constant, not a once-a-year exercise.

AI finally speaks the user's language.

Modern AI and LLMs give the natural, guiding interface that old proprietary ML and recommendation engines never could, putting expert cost intelligence in everyone's hands.

Three shifts. One window.
Apex, now.

04 · See it work

Upload a part. Get the answer.

The interface is a conversation. Type the ask in plain English; AI agents read the drawing, run the deterministic models and return the should-cost with its evidence. A week of expert work becomes an upload.

Try it on

Apex Cost · conversation

Buyer

Cost this horn assembly. Target volume 50,000 a year, from our Gurgaon supplier, landed to Detroit.

Apex

Parsed: 28 child parts, 9 process steps. Wire drives 29% of unit cost, worth a deeper look. Building the model from live Gurgaon rates.

Should-cost $1.58 / unit · confidence ±4% at 95%

Ingest

Drop in the drawing, 2D, 3D, or a torn-down part.

Understand

AI reads the dimensions, materials and tolerances straight off the drawing.

Match process

Works out how the part would actually be made: cast, machined, molded, assembled.

Model cost

Builds the cost line by line from live, local rates for that exact city and volume.

Fact pack

A signed, evidence-backed report the buyer takes into the negotiation.

3-5 days → 3-5 minutes From drawing to defensible should-cost and negotiation fact pack.

The cost stack · built bottom-up

Raw material

$0.85

Processing

$0.28

Logistics

$0.18

Overheads

$0.27

Should-cost · per unit$1.58

Apex Suggest · what to watch

Apex

Copper wire swings this the most, up to ±$0.101 a unit on a full commodity swing. Lock a hedge or index it in the PO so the next spike is shared, not absorbed. Labour rate and volume are smaller, second-order watches.

Apex flags this the moment the model is built. No extra step for the buyer.

Cost elasticity

Relax a tolerance from 0.002" to 0.005" where the function allows it. 15% cheaper, with the trade-off shown before the design is locked.

Material swap

Move from 304 stainless to powder-coated mild steel for a non-corrosive duty. 40% cheaper, same fit and function, a swap already proven by 32 comparable parts in production.

Negotiation

Three suppliers quote $12.50, $11.80 and $18.00. The should-cost says $10.20. Apex flags the $18.00 as a likely spec misread, notes the middle bidder's 95% on-time record, and sets the walk-in target: $10.80.

Why the answer can be trusted

It checks itself

If an extracted value looks impossible, a 0.001 mm wall, a negative cost, the agent flags it instead of running with it.

It asks when unsure

An ambiguous drawing gets a question back, not a guess. Missing specs are filled from similar parts and marked as assumptions.

It learns from correction

Every estimate carries a confidence range, and every expert override becomes training signal for the next one.

Experts stay in command

The workflow runs end to end on its own, and an expert can step in, override or take over at any point. Human optional, expert welcome.

Beyond the estimate

What-if scenarios

Change any assumption, volume, material, tariff, machine rate, and the cost rebuilds instantly. Cost the part in the future you are negotiating for, then save the scenario and share it.

Fact packs on record

Every decision becomes a cryptographically signed PDF with its evidence bundle: audit-ready, dispute-ready, retained for seven years. Institutional memory instead of tribal knowledge.

Multiplayer by default

Costing sessions are shared. Colleagues watch live, run their own scenario alongside, and send the finished pack straight to the CFO.

The estimate is not the product. The negotiation it wins is. Buyers do not buy numbers, they buy evidence.

05 · Cost + sourcing

Cost the part everywhere. Source it for less.

For any part, Apex builds the should-cost across regions and suppliers, then guides the best way to source it: which location, which supplier, and what it should cost you.

Landed should-cost · same part, same drawing, four regions

NingboChina

$1.56Lowest landed

GurgaonIndia

$1.58Current supplier

MonterreyMexico

$1.62

DetroitUSA

$2.32

Apex Suggest · Three paths, ranked

The cheapest bar is not the decision. The buyer picks the path.

AStay and negotiate

Effort NoneLanded cost Holds at $1.58Best for Zero risk, zero capex

Ningbo saves $0.02 a part, about $1,000 a year here, not enough to repay a five-figure requalification. Better: hold Gurgaon at $1.58, index the copper, and share the next commodity swing instead of absorbing it. No capex, no added risk.

BHedge nearshore

Effort Duplicate tool + qualificationLanded cost $1.62Best for Tariff and freight cover

Qualify Monterrey at $1.62 as a second source: USMCA duty-free, four days by truck to Detroit, and cover against tariff or freight shocks on the Asia lane. Cost: one duplicate tool, one qualification run. Apex costs both and shows the payback upfront.

CRe-engineer and reshore

Effort Engineering change, snap-fit jointLanded cost Detroit $2.32 → $1.78Best for When tariff or transit risk outweighs cost

Swap the screwed-and-crimped case joint for a snap-fit, cutting assembly labour 60%. Every region gets cheaper, but labour is the spread, so the high-wage bar falls furthest: Detroit $2.32 $1.78. It still costs more than staying, and Apex says so. It wins when tariff risk or nine weeks on the water outweighs cost, not before.

Apex Suggest costs each path, including the engineering changes, and ranks them against the program's risk posture. It is just as willing to say "stay" as "move", and the buyer always makes the call. Supplier discovery and scoring widen the field as Apex Discover comes online.

06 · What it watches

It surfaces the decision before you ask.

A should-cost is not a report, it is a living number. Apex watches the drivers underneath it and turns movement into decisions.

Commodity

Copper rate +4.8% this week.

A signed magnet-wire bid now sits below the live copper rate. Award it within seven days, before the supplier walks the number back.

Electronics

DRAM memory +18% this quarter.

AI-server demand is driving memory cost up. Re-run exposed electronics parts lists to see the per-part impact before the next award.

Tariff

Tariff review on horn duty.

Two open bids rely on a single country. Hedge across an alternate trade lane.

Quality

Supplier passed its re-audit.

Qualification cleared. Safe to award the split order.

And it scales from the part to the portfolio. A buyer who owns aluminum extrusions sees total spend, savings trending against the metal index, single-source risks and design-freeze deadlines in one view. The level a manager thinks at, not just the level an engineer works at.

07 · Platform

Focused now. Built to expand.

The wedge is cost. The platform is everything cost touches: suppliers, sourcing, and eventually carbon. Each product runs on the same data foundation, so every expansion starts ahead.

Currently under development

Hero product

Apex Cost

Deterministic should-cost models, run by AI agents from the drawing up.

Cost guidance

Apex Suggest

Learns from every estimate and surfaces the cheaper path before you ask.

Foundation · included at no extra cost

Apex Data

The live, geo-indexed cost-driver foundation every model runs on.

Shipping alongside · enterprise enablers

Integration

Apex Connect

Plugs into the systems manufacturers already run: ERP, PLM, CAD, files and data warehouses. No-code connectors that take minutes to set up, not months.

Security & compliance

Apex Trust

Built for the enterprise checklist from day one: SOC 2 and ISO 27001 readiness, GDPR, audit trails and uptime commitments in one portal.

Later on the roadmap

Planned

Apex Discover

Supplier discovery and cost scoring, widening the sourcing field.

Planned

Apex Sustain

Product carbon footprint and Scope 3, from the same cost models.

08 · The founder

A decade of cost engineering, encoded into AI.

Vaibhav Kumar

Founder · Apex Current

LinkedIn →

Vaibhav Kumar spent the last decade solving cost engineering problems for the world's largest manufacturers.

Founder · Advanced Structures India (ASI) Teardown-grade cost analysis Automotive Electronics Appliances Off-highway Industrial equipment

His first company, ASI, built that practice from the ground up. It runs today as a sustainable, professionally managed business.

Apex is his second company in this space, and a different bet. Where ASI delivers cost engineering through a technology backed expert led solution, Apex encodes that same ground truth, how products are actually designed, made, costed and procured, into a self learning platform built on AI and data at global scale.

Ten years on the ground gives Apex something else that matters just as much: a precise read on how customers will actually use the platform, which cost problems are worth fixing first, and how to fix them. That is product judgment competitors cannot buy or replicate quickly, and it is what shapes Apex from day one.

His decade of track record working with manufacturers like

GoogleAmazon RoboticsWhirlpoolStanley Black & DeckerStarbucksEnphaseToyotaHondaHyundaiVolvoJaguar Land RoverDaimler TruckValeoAisinGarrettCNHMG MotorMahindraTata ElectronicsTVSHero MotoCorpBlue StarAtherGreaves Cotton

And many more, across six industries. Apex is the earlier-stage company built on this ground truth, not the vendor of record for these relationships.