Howdy!
I’m Kyle Bedrich.

I build small aircraft and the tools to design their propulsion systems.

I’m the solo founder, maintainer, and developer of ThrustLab, and an aerospace engineering master’s student at Texas A&M.

Kyle Bedrich holding a heavy-lift octocopter
DARPA Lift Challenge

SAE Aero Design

Texas A&M · 2022–2025

  • Python
  • NumPy
  • SciPy
A Texas A&M Micro Class aircraft on the runway during testing

Chief Engineer, Micro Class

2024–2025

Electronics & Propulsion Lead

2023–20241st overall2024 SAE Aero Design West
Micro Class · team result
1st in mission1st in Flight Demonstration Readiness Review

I spent three years on Texas A&M’s SAE Aero Design team, designing small aircraft to complete very specific missions. I worked on propulsion, aircraft configuration, and flight testing, then led the Micro Class team as chief engineer.

Most importantly, I built the team’s in-house simulation software for takeoff and flight dynamics, aircraft optimization, and propulsion system design. We used it to make design decisions and plan what to carry in the conditions we actually had at the field.

Flight testing the 2024–25 Micro Class aircraft.

At competition

Our flights from the 2025 SAE Aero Design East livestream.

The 2024–25 propulsion study: a dense cloud of thrust-versus-electronics-weight results with the Pareto front colored by flight score
The original 2024–25 study. Color shows modeled flight score; red crosses mark the 20×6.2 propeller / 400KV motor candidates.

Choosing the powertrain

For 2023–24, I used PROM to compare 67,704 motor, propeller, and battery combinations. We selected for thrust-to-powertrain weight, with enough battery for a go-around.

The 2024–25 study compared 35,980 combinations. I pulled out the Pareto front, then ran those candidates through KTOC to see which produced the best flight score.

More thrust is useful, but the extra weight has to earn its place on the aircraft.

KTOC

Kyle’s Takeoff Code. I built it to work out how much we could carry and when to rotate. It models the elevator, servo speed, wheel loading, and the landing gear’s effect on rotation. It also has a rotation-timing optimizer, optional ground effect, and climb-to-cruise simulation.

Takeoff weight and headwind

MCP · 2024–25
Within rotation limitInfeasible: rotation >18°
0510152025456789Takeoff distance · ftTakeoff weight · lb
6.2 lb
7 mph
Modeled takeoff distance
8.5 ft
Rotation at liftoff
16.8°

Within the 18° rotation limit.

Original KTOC model: 5° flaps, 450 W powertrain, 0.5 s rotation delay. Air density, geometry and pitch inertia stay fixed. Feasibility here checks the 18° rotation limit.

FlightLine

FlightLine takes that work into a full mission. I can link takeoff, climb, cruise, and turns, with an aircraft model and propulsion system carried through the whole flight.

It includes sustained wind, a Dryden gust model, control-surface limits, and state histories for plotting. This mission uses the earlier 5DoF model. I also worked on a 6DoF version with LQR control.

Original Python 3D plot from FlightLine’s three-lap demonstration mission, showing takeoff, climb and repeated turns
3 lap demonstration mission in FlightLine.
Kyle with two of his team’s winning aircraft
Two of my team’s winning aircraft.
The Texas A&M SAE Aero Design team at an award ceremony
SAE Aero Design awards ceremony.

DARPA Lift Challenge

2025–present

  • SOLIDWORKS
  • Python
  • NumPy

The 2026 DARPA Lift Challenge was a heavy-lift drone competition scored by payload-to-aircraft-weight ratio. The course covered four nautical miles carrying payload and one unloaded.

Our aircraft could carry 110 lb for four nautical miles, with a 2.55:1 payload-to-aircraft-weight ratio.

Our flight at the DARPA Lift Challenge.Watch from 37:39 on YouTube
Studio render of the fully assembled DARPA octocopter, viewed from above at an angleStudio render of the fully assembled DARPA octocopter, viewed from above at an angle
DARPA aircraft landing gear
Landing gear.

Propulsion & flight testing

I work on propulsion and configuration selection for our heavy-lift aircraft, along with flight-test planning.

The motor, propeller, and battery have to work together through the flight. I built analysis tools to compare those combinations and look at how much throttle is needed as the battery discharges.

Propulsion selection

Six-motor study
050100150200250405060708090100lbfThrottle · %
65 %
Total thrust
111.0 lbf
Electrical power
6.76 kW

T-Motor 8017 120KV, six motors, 48 V. Manufacturer table points from my DARPA propulsion workbook; static estimates without rotor-interaction losses. This study is separate from the octocopter pictured above.

Hydrofoil structural design

AERO 604 · Team project · 2025

  • Abaqus
  • Python
  • SciPy
The hydrofoil assembly modeled in Abaqus, including the board, mast, fuselage and lifting surfaces
The hydrofoil model in Abaqus.

Automating the structural design

I automated Abaqus with Python to build, mesh, load, and evaluate the structure. I used topology optimization to develop the ribs and mast spar, modeled carbon-fiber layups, and connected the structural model to an optimizer.

We varied seven design parameters across the board, mast, and fuselage. Each run checked mass, deflection, stress, and buckling. A Bayesian optimizer chose the next design to evaluate.

Final model mass
10.0 kg
Board deflection
5.45 mm

Results from the final structural analysis.

Abaqus topology study of the hydrofoil board's internal support, with material removed between the load paths
Board support: topology optimization helped define the internal load paths.
The parameterized hydrofoil mast and board support in the Abaqus structural model
The mast spar after topology exploration and parameterization.
Stress contours from the optimized hydrofoil structure in Abaqus
Stress in the final configuration under the front-heavy load case.
The critical buckling mode of the optimized hydrofoil, concentrated in the composite board
The critical buckling mode. Deformation is scaled for visibility.

ThrustLab

Solo founder & developer

  • Python
  • Rust
  • TypeScript
ThrustLab motor and propeller

It started with the SAE team.

The propulsion model began as a simple exercise of first principles. Over 30 versions were used by me and my peers to design our aircraft. I kept developing it into ThrustLab.

A proper engineering process requires models to be validated so that the engineer using the model knows where to trust it. That is why I created ThrustLab.

PROM

Motor, propeller, and battery models for comparing propulsion systems. The original SAE code became the foundation for ThrustLab.

KTOC & FlightLine

From takeoff and rotation to complete aircraft missions. KTOC grew from the team’s earlier STOIC work; FlightLine extended the mission simulation.

Aircraft design tools

Airfoil and wing optimization, scoring sensitivity, and configuration trade studies used during the SAE design process.

Built with

Frontend
  • TypeScript
  • React
  • Next.js
  • Tailwind CSS
  • Plotly
  • Three.js
Backend
  • Python
  • FastAPI
  • Pydantic
  • SQLAlchemy
  • PostgreSQL
  • Redis
  • Celery
  • Django
  • Wagtail

Django and Wagtail manage the site’s content.

Solver
  • Rust
  • PyO3

Rust connected to Python through PyO3. Coupled electrical, aerodynamic, and thermal models.

Infrastructure
  • Hetzner
  • Docker Compose
  • Cloudflare Tunnel

Three servers with separate web, solver, and database roles.

Traffic
  • Cloudflare Tunnel
Web server
  • Next.js
  • FastAPI
  • Redis
App, API, and job queue
Solver server
  • Celery
  • Rust
Background simulations
Database server
  • PostgreSQL
  • PgBouncer
Shared by the web and solver services

Simulations run in background workers so long solves don’t tie up web requests. The platform also has a REST API, a Python SDK, and FMU export for connecting propulsion models to other simulation tools.

Hosted in Falkenstein, Germany

ThrustLab’s web, solver, and database servers run in Hetzner’s Falkenstein data center park. Hetzner sources 100% renewable electricity for its German data centers, backed by renewable-energy Guarantees of Origin.

Hetzner reports an average data-center PUE of 1.13—about 0.13 kWh for cooling and other facility systems per 1 kWh used by IT equipment. Outside-air cooling and efficient power distribution help keep that overhead low.

The Texas A&M SAE Aero Design team with their aircraft at the field

A little about me.

I’m based in College Station, Texas. I graduated with my bachelor’s in electrical engineering in 2025 and am currently getting my master’s in aerospace engineering at Texas A&M.

I learned aircraft design on the SAE team, through systems design, CFD, FEA, and a lot of building and testing.

Get in touch.

I’m looking for a full-time opportunity after I graduate with my master’s in aerospace engineering in December 2026.

If you want to talk aircraft, propulsion, or work together, send me an email.