VENKAT TERLI

SCROLL TO EXPLORE ↓

AVAILABLE FOR WORK

( 01 ) SELECTED RESEARCH

RESEARCH IN MOTION,
DETAILS UNDER WRAPS

UNPUBLISHED WORK: SUMMARIES KEPT INTENTIONALLY HIGH-LEVEL. ASK ME FOR THE FULL STORY.

01 Certifiable Orbital RL Competed at TNJSF 2026 In Progress Learning-based spacecraft controllers for multi-body gravitational environments, with an emphasis on recovery regions, stability-aware objectives, and certificates that describe when a learned controller can be trusted. With Eshan Kulkarni, bound for the 2027 ISEF season. ✦ EARLIER ITERATION AT TNJSF 2026: 2ND PLACE, PHYSICS & ASTRONOMY · AIR FORCE RESEARCH LAB AWARD · CITADEL SECURITIES INNOVATION PRIZE · INNOVATION AWARD DEEP RL · ASTRODYNAMICS 2026 02 Where Surrogates Fail Draft Complete Private Manuscript A scientific ML project studying when fast surrogate models appear locally accurate while still producing misleading global error signals. Current work focuses on diagnostic rankings for PDE models and visual tools for understanding where surrogate predictions fail. SCIENTIFIC ML · UQ · PDEs 2026 03 Covenant Prototype In Progress Public Demo Soon A Python library that lets compound AI pipelines make promises about output types, reliability, and resource budgets, and keep them. Each stage declares a contract, composition derives the combined guarantee automatically, and the runtime refuses to let a pipeline silently break its word. AI RELIABILITY · LANGUAGE DESIGN 2026

( 02 ) PLAYGROUND

SMALL MACHINES,
BIG FEELINGS

Prototype

P.01

Reaction–Diffusion

GRAY-SCOTT · LIVE SIMULATION

Prototype

P.02

Boids

FLOCKING · EMERGENT BEHAVIOUR

Prototype

P.03

Wavefunction Collapse

CONSTRAINT SOLVING · TILES

Prototype

P.04

Fourier Doodles

EPICYCLES · COMPLEX SERIES

Prototype

P.05

Cellular Morphogenesis

GROWTH · REGENERATION

Prototype

P.06

Double Pendulum Choir

CHAOS · SENSITIVE DEPENDENCE

( 03 ) ABOUT

I train agents, prove things about them, and build playgrounds where the mathematics becomes something you can touch.

Currently: teaching spacecraft to fly themselves with provable stability, teaching surrogate models to admit where they're wrong, and teaching AI pipelines to keep their promises. I believe a proof and a particle system are the same instinct: both are ways of asking a structure to reveal itself.

0RESEARCH PROJECTS IN FLIGHT
0TNJSF 2026 AWARDS
0PLAYGROUND EXPERIMENTS
0THEMES ON THIS SITE

01 / RESEARCH

Formulate, prove, ablate. I like problems where the math pushes back: reward design, convergence, emergent behaviour.

02 / ENGINEERING

From CUDA kernels to browser canvases. Systems that train fast, run anywhere, and fail loudly.

03 / ART

Simulation as a medium. I turn value functions, vector fields and eigenmodes into things you can feel.

RECOGNITION: TERRA NORTH JERSEY STEM FAIR 2026 (ISEF QUALIFIER)

2nd Place, Physics & Astronomy CATEGORY AWARD · PH.03 · W/ ESHAN KULKARNI For “Leveraging Deep Reinforcement Learning for Trajectory Optimization and Fuel Efficiency in N-Body Systems”, the first iteration of the certifiable orbital RL work above, at a fair that sends 7–8 projects to the Regeneron ISEF each year.
Air Force Research Laboratory Award ONE OF 4 PROJECTS FAIR-WIDE · RESEARCH OF INTEREST TO THE AIR FORCE
Citadel Securities Innovation Prize ONE PROJECT FAIR-WIDE · INGENUITY & INNOVATION
Innovation Award RESOURCEFUL EXPERIMENTATION WITH ACCESSIBLE TOOLS