B.S. Aerospace Engineering · Rutgers University · Class of 2028

Anudeep
Mantravadi

Seeking a Summer 2027 engineering internship

Skills

Design & simulation
SolidWorks, Siemens NX, GD&T, AutoCAD, ANSYS, OpenFOAM
Manufacturing & hardware
Wire arc additive manufacturing, CNC, machine shop fabrication, 3D printing, DFM/DFA
Programming
Python, MATLAB, C/C++, PyTorch, Linux

Leadership

  • Aerospace Structures Deputy Lead, Rutgers Airborne (AIAA). Led an 18-engineer structures sub-team from requirements to flight test.
  • Leg Mechanisms Lead, NASA L'SPACE NPWEE. Led the leg subteam and CAD; named co-inventor on a NASA invention disclosure.
  • Mechanical Engineering Lead, RU HART. Leads a 12-engineer mechanical sub-team building Max Bot 2.0, an autonomous food delivery robot.
  • CNC vertical milling certified by Rutgers University.

Micro Wire Arc Additive Manufacturing Machine

My Role Designer, Builder, Wiring & Controls Lead

A 3D-printer-scale wire arc additive manufacturing machine built from scratch into a working deposition system. I designed the 170+ part gantry in SolidWorks, machined and assembled the moving parts, wired four RAMPS driver boards by hand, and tuned the wire feed to stay consistent through a full deposition.

Funded by National Science Foundation Rutgers Department of Mechanical and Aerospace Engineering

National Science Foundation and the Rutgers Department of Mechanical and Aerospace Engineering

Loading assembly...
Real CAD model Drag to rotate, scroll to zoom
Live deposition pass: the arc melts the wire onto the substrate as the gantry moves.
Micro WAAM gantry assembly, wide view
The assembled 3-axis gantry, with each stepper calibrated against vernier measurements before the deposition head went on.
Z-axis and wire feed mechanism detail
Z-axis lead screw checked for load and buckling, and spool position set with wire-sag analysis so live wire never touches the frame.
Arc deposition closeup with sparks
Controlled deposition tests tracked with GD&T in AutoCAD, plus a Python tool that sorts bead cross-sections into nominal, overprint, and underprint for the lab's ML work.
3D printed wire feed extruder mechanism
Printed wire-feed housing with guides and a spring-loaded pinch roller, tuned until feed stayed steady through a full deposition.
Custom wiring harness across RAMPS boards
Four RAMPS boards (X, Y, Z, extruder) wired and labeled by hand, each motor isolated and tested before integration.

NASA Lucy Mission, L'SPACE NPWEE

My Role Leg Mechanisms Lead · Lead CAD Designer · Team 18

Ten students and early-career engineers from eight universities wrote and costed a NASA Space Technology Mission Directorate proposal through the L'SPACE NASA Proposal Writing and Evaluation Experience (NPWEE), part of NASA's Lucy mission student pipeline. I led the leg mechanism subteam and the CAD, and was named a co-inventor on the team's NASA invention disclosure. Project details are restricted, so only outcomes are shown here.

Program NASA L'SPACE

NASA, through the L'SPACE program and the NASA Proposal Writing and Evaluation Experience (NPWEE)

NASA Form 1679 Disclosure of Invention and New Technology naming Anudeep Mantravadi as innovator, with co-innovator details blurred
NASA Form 1679, Disclosure of Invention and New Technology. Co-innovator details blurred, technology title withheld.
01

Co-invented a power-free landing leg technology featured in an official NASA New Technology Report (invention disclosure, routed to NASA Marshall Space Flight Center), credited as a named innovator for the detailed CAD design.

02

Targeted 10+ landings per leg versus 1 for Apollo-style single-use struts.

Blurred CAD of the lander leg system, marked ITAR-restricted intellectual property per NASA
Leg mechanism CAD (Siemens NX and Fusion 360). Blurred: ITAR-restricted intellectual property per NASA.
01

Raised projected tip-over margin 56% (21.6° to 33.8°) on a 20° slope by modeling the lander in Siemens NX and Fusion 360 CAD.

02

Delivered a drop-test-ready lander design 5.5% under NASA's $10K budget ($9,450 estimated build) by detailing the leg mechanism CAD and drawings.

CFD Validation and Physics-Informed Neural Networks

My Role CFD Analyst, ML Engineer

Two linked tracks. First, RANS CFD validation of a Clark Y airfoil in OpenFOAM against XFOIL. Second, a forward physics-informed neural network (PyTorch) that solves unsteady incompressible flow at Re=100 with no CFD training data. Inferring viscosity from sparse sensors with an inverse PINN is the next phase.

Funded by Rutgers Department of Mechanical and Aerospace Engineering

Rutgers Department of Mechanical and Aerospace Engineering

Gmsh structured mesh around Clark Y airfoil with boundary layer refinement

Gmsh mesh of about 65,000 cells around a Clark Y airfoil at 15 degrees angle of attack, refined near the airfoil and across the wake, built from scratch.

Custom boundary condition verification script output
check_patches.py flags mislabeled boundary patches before a solver run, written after a swapped inlet and outlet wasted an early case.
simpleFoam convergence and force coefficient output
simpleFoam converged in 2,403 iterations before any force was read. Across 34 cases, lift landed within about 3% of XFOIL.
PINN training log and predicted velocity and pressure fields after 2,000 epochs
Early PINN test on flow past a cylinder: total loss fell from 1.2 to 0.0007 over 2,000 epochs with no CFD training data.
PINN versus OpenFOAM velocity and pressure field comparison scatter plot
Clark Y PINN versus OpenFOAM at Re=100: this snapshot shows L2 error of 25% in u, 84% in v, and 160% in p. Fourier features and airfoil-interior exclusion later cut combined error 74% from baseline.
Ongoing

Open items: the drag mismatch against XFOIL, boundary-effect propagation, a better PINN loss function, and data density. The inverse (viscosity) PINN follows once the forward model is solid.

AIAA Design-Build-Fly: RU Airborne Aircraft

My Role Aerospace Structures Deputy Lead

Rutgers Airborne's entry in AIAA Design-Build-Fly. Constraint and sensitivity analysis sized the airframe, and the aircraft was built and flown. As structures deputy lead I led an 18-engineer sub-team from requirements to flight test; my own CAD and FEA covered the wing spar, firewall, and landing gear.

Team Rutgers Airborne AIAA

Rutgers Airborne, competing in the AIAA Design-Build-Fly competition

The flown Design-Build-Fly aircraft. I sized it with constraint and sensitivity analysis and ran FEA on the primary load paths before fabrication.

Full isometric exploded CAD view of the aircraft with 37 labeled components
Full SolidWorks model, 37 labeled components; my CAD work centered on the wing spar, firewall, and landing gear.
Constraint analysis plot for thrust-to-weight versus wing loading
Constraint analysis: takeoff, cruise speed, and turn load factor plotted together to fix thrust-to-weight and wing loading.
Sensitivity analysis plot for mission score versus design parameters
Sensitivity analysis showed banner length and bank angle dominate mission score, so design effort went there first.
Wing spar finite element analysis stress distribution
Wing spar FEA: von Mises stress with the half-wing modeled as a fixed beam under gravity.
Main landing gear finite element analysis deflection
Landing gear FEA under worst-case landing load, with deflection checked against clearance. The aluminum redesign reclaimed 1.0 lb of payload.
Firewall finite element analysis strain distribution
Firewall strain map around the motor mounts, used to remove weight without weakening the mount points.

Autonomous Inspection Robot on Boston Dynamics Spot

My Role Mechanical Design & Systems Integration

A 7-subsystem remote inspection robot on a Boston Dynamics Spot: scissor-lift sensor mast, pan-tilt thermal/RGB head, towed Niryo Ned2 arm cart, and a Jetson Orin Nano controller. Four USB cameras, a thermal camera, and temperature/humidity sensors feed a Gemini vision-language model and a UI dashboard that flags heat sources and humidity spikes.

Funded by Boston Dynamics Rutgers Department of Industrial and Systems Engineering

Boston Dynamics, which sponsored the $75,000 Spot robot, and the Rutgers Department of Industrial and Systems Engineering, which funds the project

Spot inspection robot with scissor-lift sensor mast and towed Niryo Ned2 arm cart
Spot with the scissor-lift sensor mast, pan-tilt thermal/RGB head, and towed Ned2 arm cart.
Niryo Ned2 arm on 3D-printed cart mount
Niryo Ned2 arm on its 3D-printed cart mount, towed behind Spot to retrieve objects remotely.
3D-printed cart deck with locating features
Printed cart deck with locating features, test-fitted before final assembly.

BoneLoad, Astronaut Bone-Loading Monitor

My Role ML Model Training & Development · Team BioSensors
QR code linking to the live BoneLoad demo Live demo Try BoneLoad yourself Open the real-time monitor in your browser. No hardware needed: choose the Mock generator source. Scan the code to open it on your phone. Launch demo ↗

Built at the 2026 NASA Space Apps Challenge hosted by NYU Tandon, answering the "Create Exercise Assist Software for Astronauts" challenge. Astronauts lose roughly 1.5% of bone density per month in microgravity, so BoneLoad asks one question each day: was today's loading enough? Three body-worn accelerometers stream to an ESP32-C3 and a real-time web monitor.

Event NASA International Space Apps Challenge NYU Tandon School of Engineering

NASA International Space Apps Challenge, hosted by NYU Tandon School of Engineering

BoneLoad real-time monitor showing live leg model, bone load per segment, joint angles, and key forces

Live monitor: a 3D leg model with per-segment bone load, joint angles, and estimated forces at about 9 Hz. A k-nearest-neighbors classifier (rest, walk, squat) reaches 97.8% balanced accuracy on NASA bed-rest data plus team recordings.

Wearable sensor rig with three accelerometers wired to an ESP32-C3

Wearable rig: belt, mid-thigh, and below-knee accelerometers on the ESP32-C3, with neutral-pose calibration turning segment tilt into joint angles.

LUNARIS: Lunar Autonomous Navigation and Resource Identification System

My Role Pipeline Architect, Sole Developer

Water ice at the lunar south pole could become propellant, but choosing where to mine means fusing data from several NASA instruments by hand. LUNARIS is an open-source pipeline I built to automate that: from raw NASA orbital data to ranked candidate sites with economic valuations in under 90 seconds. It has five layers: terrain intelligence, ice probability mapping, multi-objective ISRU scoring, rover coordination, and a Moon-to-Mars transfer experiment.

Built for Rutgers Innovation Awards

Rutgers Innovation Awards

ISRU Site Selection Heatmap
Terrain safety, ice probability, and solar access combined into one ISRU score that ranks sites by value per square kilometer.
A* Pathfinding on Lunar Terrain
A* routing over the CNN terrain cost map toward the highest-value sites.
Rover Swarm Coordination Simulation
Rover swarm simulation: several rovers share one hazard map without converging on the same path.
DANN Moon to Mars transfer learning training curves
Moon-to-Mars transfer experiment: source accuracy reached the high 80s but Mars accuracy stayed at roughly 40 to 60%, so transfer is unsolved. Shown as is.

Submitted to the Rutgers Innovation Awards 2026 and NASA RASC-AL 2027.

Height-Adjustable Desk

My Role Mechanical Design, Motion Analysis

A height-adjustable desk modeled in SolidWorks around two scissor lifts driven by a linear actuator. Every fastener, bearing, and weld bead is modeled, in sheet metal and purchased hardware.

SolidWorks motion study through the full contracted-to-extended range, checking the lift stays level and collision-free.
Kinematics Profile
Linkage checked in contracted and extended positions for pivot spacing, platform leveling, and actuator travel.
Engineering Drawing
Part drawings dimensioned for the lead-screw and bearing fits, then cross-checked against the assembly before release.
Exploded Assembly
Exploded view showing build order and confirming fastener access at every joint.

Rotating Detonation Engine, Cold Flow Test Rig

My Role Design, Instrumentation & Data Pipeline Lead

A small-scale cold-flow rig for studying rotating detonation engine wave dynamics without combustion. Compressed air passes through 12 injection holes angled at 45 degrees, which add tangential momentum and set up a rotating pressure wave in the annulus. Two MPU6050 accelerometer/gyro modules on the outer wall, one above the converging section and one below the throat, record the response from 30 to 90 PSI.

Full cold-flow test rig with regulated air supply and instrumentation

The cold-flow rig: regulated 30 to 90 PSI air supply with wireless pressure readout, a printed converging-diverging housing, and dual MPU6050 sensors logged by a Raspberry Pi Zero.

Onshape CAD model of the RDE housing with construction planes for the angled injection hole pattern
Onshape model with a 12-plane radial pattern setting each injection hole at 45 degrees; symmetry checked before printing.
3D printed RDE housing showing the converging-diverging bore and angled injection holes
Printed housing with the converging-diverging bore and angled injection holes at the rim.
Dual MPU6050 sensor wiring on the printed housing
Two MPU6050 modules on one logging board, above the converging section and below the throat, to compare the wave signature across the geometry.
CFD streamlines through the injector ring, showing the rotation the angled holes impart.
CFD turbulent kinematic viscosity contour through the injection holes and channel
Turbulent viscosity contour from the same CFD, showing where flow through the holes is still developing.
Flask dashboard with live FFT and health metrics. An Isolation Forest (10% contamination) screens the 6,792 logged samples for outliers, and a vibration-based turbulence index peaked at 30 PSI.
RDE injector plate top view with GD&T callouts and datum references
Injector plate, 12 ports, with concentricity and runout callouts on a dual-datum scheme.
RDE nozzle cone axial cross-section with converging and diverging dimensions
Nozzle cone drawing for the hot-fire design (Inconel 718); the cold-flow rig uses a printed version.
Tolerance stack-up analysis table comparing worst-case and RSS methods against design limit
Tolerance stack-up, worst-case versus RSS, with 8.3% design margin remaining.
Engineering pipeline

GD&T drawings for the injector plate, nozzle cone, and assembly carry worst-case stack-up analysis. The SPC control plan flagged throat diameter at Cpk 1.11, below the 1.33 target, as the dimension most likely to drift across a print run.

Adjustable Swept Wing and Wind Tunnel Test System

My Role System Design, Electronics & Data Acquisition

A small low-speed wind tunnel for testing a NACA 4415 airfoil. An Elegoo UNO R3 drives two synchronized MG90S servos to set angle of attack, and a BMP280 near 52 percent chord on the upper surface reads static pressure. Lift and drag coefficients are inferred from that single pressure reading, not measured with a force balance, so results are best read as trends.

Tunnel with an adjustable NACA 4415 wing: two synchronized servos set angle of attack while the BMP280 reads surface pressure.

Airfoil Test Section
Test section and sensor mount, with results shown live on the Serial Monitor and an ST7735 TFT during each sweep.
Stall Behavior & AoA Sweep
Inferred CL, CD, and L/D against angle of attack from the sweep, showing where the wing stalls.
Velocity Effects on Aerodynamic Performance
Lift force and upper-surface gauge pressure against tunnel velocity at several fan gear settings.
Drag Polar Curve
Low-Reynolds-number drag polar from the inferred coefficients, used to find the angle-of-attack range where lift-to-drag peaks.

Let's build
something
real.