Designing systems that work on paper — and hold up in the real world.
From CNC fixtures to walking robots to EV enclosures — built to be measurably better.
Hi, I'm Prabdeep — a mechanical engineer who lives at the intersection of design and the shop floor.
I started as an intern in a fabrication shop, learning fast that a beautiful CAD model means nothing if machinists can't build it — or if it falls apart after 200 hours in the field. That gap between engineering intent and real-world execution is exactly where I operate.
From 6-legged robots to industrial conveyors to EV battery enclosures — I design with manufacturing in mind, validate with FEA and field data, and don't call something done until the numbers prove it.
Currently finishing my MS at UT Dallas (May 2026). Based in Dallas, TX. Open to full-time roles across the US.
18-servo walking robot. 120 parts. ±0.003″ tolerance across 18 CNC-machined components. Built from scratch, validated to 800 hours MTBF. The first version didn't work — here's what I did about it.
Read Case Study →Autonomous ground platform for rough agricultural terrain. Designed for real dirt, not a lab. 58% fewer weld joints, 35% better torsional rigidity, 650+ hours MTBF after 40+ hours of field trials.
Read Case Study →15+ industrial assemblies engineered end-to-end. Zero clogging failures across 2,400 operating hours. 40+ GD&T CAD models that cut shop-floor questions by 85% and eliminated $12K in rework.
Read Case Study →Structural & thermal FEA validation under 3G crash loads. <2mm deflection. 15°C thermal limit. $500K prototype program with 95% design reuse carried into production.
Read Case Study →I'm finishing my MS at UT Dallas (May 2026) and open to full-time mechanical engineering roles across the US. If you have a hard problem that needs a data-driven engineer who bridges design and the shop floor — let's talk.
A six-legged walking robot sounds cool on paper. Getting 18 servos, 120 parts, and 18 CNC-machined components to move in perfect coordination — without binding, without failure, without constant rework — that's where the real engineering starts.
This was my project. And the first version didn't work.
Most robotics projects fail quietly in the tolerance gaps. A leg joint that's 0.008″ off doesn't look wrong. But multiply that across 6 legs, 3 joints each, and suddenly your robot walks with a limp — or seizes under load.
We had 6 tolerance stack-up issues in the first build. Components that looked fine individually created interference when assembled. The robot couldn't complete a full gait cycle without mechanical binding. I could have iterated. Instead, I went back to the drawings.
I audited every mating interface across all 18 machined components. Tightened critical fits to ±0.003″, rebuilt the tolerance chain from the hip joint outward, and worked directly with the machinist on 3 components that needed recuts.
Then I ran FEA and FMEA on all 24 structural components — not just looking for stress concentrations, but specifically hunting for failure modes that don't show up until hour 150 of operation. I found 5. Fixed them before a single part was touched.
I standardized the entire BOM with step-by-step work instructions for all 12 sub-assemblies — so the build process became repeatable, not dependent on whoever happened to be assembling that day.
| Metric | Result |
|---|---|
| Tolerance stack-up issues | 6 → 0 |
| Assembly iterations needed | 3 → 1 |
| First-pass acceptance rate | 96% |
| Predicted MTBF | 200 → 800 hrs |
| Positional accuracy | ±5mm / 500mm workspace |
| Repeatable build time | 4 hours, zero fit issues |
"The part I'm most proud of? The documentation. Anyone could pick up those work instructions and build it again — and it would come out the same way every time. That's engineering that scales."
Most machines look great at a demo. Then they go into the field — and real terrain, real vibration, and real weather expose every shortcut taken during design.
This project was about building something that wouldn't fall apart. An autonomous mobile platform for agricultural use, designed to handle rough ground, operate without supervision, and survive the kind of abuse that happens when nobody's watching. The benchmark we set ourselves: 650+ hours MTBF. Not a specification someone handed us — that's what we decided a reliable agricultural machine actually needed.
The original chassis had too many weld joints. Good for rigidity on paper. Bad for a fabrication shop with real labor costs and real timelines. I applied DFM systematically — not as a checkbox, but as a constraint that shaped every decision. The question wasn't just "does this work structurally?" It was "can a welder do this efficiently, and what happens if the joint isn't perfect?"
Result: 58% fewer weld joints. Fabrication time dropped 45%. $240 saved in labor on the first build alone.
With fewer joints, the challenge became maintaining structural integrity under dynamic loads. I ran FEA on the chassis layout, targeting torsional stiffness as the primary metric. Tested 4 configurations before landing on one that improved torsional rigidity by 35% while cutting weight from 45 to 40kg. Lighter. Stronger. Cheaper to build.
We ran 40+ hours of field trials across real terrain and tracked every anomaly — vibration patterns simulation didn't predict, load transfer behaviors that only appear at specific speeds. We implemented 7 targeted improvements. Not guesses. Measurements.
| Metric | Result |
|---|---|
| Weld joint reduction | −58% |
| Fabrication time | −45% |
| Labor cost saved (first build) | $240 |
| Torsional rigidity via FEA | +35% |
| Weight reduction | 45 → 40 kg (−12%) |
| Field reliability | 78% → 94% |
| MTBF achieved | 650+ hours |
"A chassis isn't done when the FEA converges. It's done when the field data says it's done."
Industrial machinery doesn't forgive ambiguity. A drawing that's unclear costs hours of back-and-forth, rework, and scrap. At ANG Enterprise, I was responsible for the full design lifecycle of ACM mill assemblies, conveyors, and ribbon blenders — from first sketch to final installation.
With a high-volume environment, every design decision had downstream consequences. I engineered 15+ ACM mill assemblies using SolidWorks with rigorous GD&T callouts, reducing redesign cycles from 3 to 1 — a 67% improvement in design efficiency.
I developed 40+ CAD assemblies with full GD&T specifically designed to eliminate shop-floor ambiguity. Clarification requests dropped by 85% and rework costs fell by $12K.
I performed FEA on 25+ shaft assemblies before any metal was cut, identifying 12 interferences that would have caused failures at the build stage. That upfront analysis prevented $8K in scrap costs and kept the production schedule intact.
Beyond design, I identified 5 production bottlenecks reducing cycle time by 18% (6.2 → 5.1 hrs) and supervised installation of 8 systems achieving ±0.005″ alignment and 87% uptime.
| Metric | Result |
|---|---|
| ACM mill assemblies engineered | 15+ |
| Redesign cycle reduction | 3 → 1 (−67%) |
| Shop-floor clarifications reduced | −85% |
| Rework cost savings (GD&T) | $12K |
| Pre-build interferences caught | 12 (preventing $8K scrap) |
| Clogging failures across 2,400 hrs | 0 |
| Scrap rate improvement | 8.5% → 2.1% |
Battery enclosures in electric vehicles aren't just structural components — they're safety-critical assemblies that need to contain high-voltage systems under crash conditions, thermal stress, and vibration. At Ward Wizard Innovations, I was responsible for validating these designs before they went into physical prototype.
Using Abaqus FEA, I modeled the battery enclosure under worst-case loading: 3G crash loads applied in multiple directions, plus thermal analysis to confirm behavior within the 15°C operational limit. The simulation had to prove the enclosure would deform less than 2mm under maximum load — without any failure at mounting interfaces.
I designed 6 EV assemblies in total, including the battery enclosure, structural brackets, and mounting hardware. Every design decision was validated against simulation before committing to tooling.
What made this project different was the emphasis on design reuse. The prototype program had a $500K budget — meaning every component that could carry forward into production directly reduced NPI costs. We achieved 95% design reuse from prototype to production.
I also implemented 14 Engineering Change Orders (ECOs) via ERP, improving BOM accuracy from 87% to 92% and reducing procurement errors by 80%.
| Metric | Result |
|---|---|
| Max deflection under 3G crash | <2mm ✓ |
| Thermal limit maintained | 15°C ✓ |
| Prototype program value | $500K |
| Design reuse in production | 95% |
| ECOs implemented | 14 |
| BOM accuracy improvement | 87% → 92% |
| Procurement error reduction | −80% |