diff options
author | Holden Rohrer <hr@hrhr.dev> | 2022-08-16 18:23:05 -0400 |
---|---|---|
committer | Holden Rohrer <hr@hrhr.dev> | 2022-08-16 18:23:05 -0400 |
commit | cf35f9ab5f78f64c62b9ae9d83df8524be58394f (patch) | |
tree | 03daec3b3d818d22c1e656b9df497f31de165a50 /20220816research.txt |
Initial Commit (research opportunities)
Diffstat (limited to '20220816research.txt')
-rw-r--r-- | 20220816research.txt | 239 |
1 files changed, 239 insertions, 0 deletions
diff --git a/20220816research.txt b/20220816research.txt new file mode 100644 index 0000000..65208f2 --- /dev/null +++ b/20220816research.txt @@ -0,0 +1,239 @@ +Niti's Research Prof: frank.dellaert@cc.gatech.edu + +Do Undergrad Research if: +- Want more depth in a specific area not covered in class +- Want to look at research for PhD + +Don't Undergrad Research if: +- You want certain results +- Want individual attention from faculty +- Want recommendations for jobs (faculty recs don't go very far in + industry anyway) + +That fits me! + +Course Reg Numbers: [2/4]69[8/9] +2: fresh/soph +4: jun/sen +8: paid +9: unpaid +|Sometimes not required but highly recommended + +Meets the junior design requirement +- LMC 4701/2 +- 9 credits of undergrad research (2 sem unpaid iirc) + +BS/MS apply after 60 credits; can do research as project/course option + +Undergrad research can be "academic" or "implementation-oriented." +- Much less grant- and paper-oriented than Masters/grad research. + +# Hyesoon Kim +## Computer Architecture +Working on GPU processor design +- CUDA Everywhere +- Open Source Design +- CuPBoP?? + +Looking for C/C++ skill set +GPU (CUDA/OpenCL) programmer +LLVM experience hopefully +Interested in compiler, runtime system + +# Frank Li +## Data-Driven Methods for Improving Internet Security + +BEES Lab. +Applies "empirical approaches" to Internet security concerns. +Probably a solutions-y thinktank? +- Large-scale Internet/Web Measurements + - Ex: massscan IPv4 range (monitor server mgmt behaviors) + - Develop new techniques for IPv6 internet +- Also data mines auth mechanisms & password leaks +- Analyze open-source Git repos to understand secure software + development +- Characterize passive DNS data to understand user network traffic + patterns + +Doesn't feel like it's for me. Too much of a social-science study + +- Qualitative User Studies + - Understand how sysadmins manage software updates + - Understand webadmins password authentication policies + - How users react to security notices/warnings + +Expects background in web security I think? + +# Rogues Gallery VIP Team +## What? +Transistor scaling/hitting the end of Moore's law. +Investigates new tech for continuing to improve performance + +"Rogues Gallery Testbed" (FPGA & FPAA) +- Unique, "rogue" hardware that needs to be programmed into useful + applications---unlike commodity memory/processors + +Worked on: +neuromorphic something +emu chick neo-memory +quantum computing +maybe reconfigurable computing + +Needs skills: +- Min: knowledge of linux, ssh, programming (C/C++/Pyth), lin algebra +- Preferred: ECE 3400 (FPAA), CX 4220, and another class I haven't taken + +# Karen Feigh +## UROC Opportunities with Helicopters + +- Cognitive Engineering + - What if we designed training and technology with how people + actually do their work? +- procedures and displays for air and space +- basic research for mil command & control decisionmaking +- adaptive intelligence for industrial robots & autonomous vehicles +- Examine human-machine interaction using field work, math, etc. + +Hiring into projects (multiple year projects getting started): +- Humans & AI Piloted Helicopter (special forces/medevac) +- Helicopter Multi-modal Symbology +Working on better AI/better human interfacing + +Requesting: +- Experience w/ unreal engine + +# Wei Xu +## NLP Research Lab + +NLP side +- Text generation +- Human-centered eval +- Multilingual NLP +- Stylistics + +Machine Learning +- Sequence-to-sequence models +- Controllability of neural network +- Cross-lingual transfer learning +- Zero-shot/few-shot learning +- Robust domain adaptation + +Want possible multilingual, interested in linguistics. +Didn't mention ML proficiency already. + +seq2seq: simplify english +Social Media Data: like hashtag word analysis + +# Anirudh Rama (or the actual prof?) +## Community wireless networks/organic mesh networks + +- Affordable internet connection +- Networks hosting local services +- Disaster scenarios + +Studying: +- Service performance/putting servers on mesh +- Handling unpredictable network topology like wireless links, + obstructions, unreliable nodes, difficult issue diagnosis + +Setting up a mesh testbed on GT campus (wireless router/antenna on +buildings; monitor perf; video conference server) + +Preferred: CS 2200, Linux SSH, Basic Networking + +# Vivek Sarkar +## Habanero Extreme Scale Software Research Lab + +- Extreme parallelism/heterogeneity at the end of Moore's Law +- Distributed parallelism, new software foundation of compilers, + runtimes, debuggers, programming languages, etc. +- Parallelism cliffs are very hard for something like Python code. + +!! See orthogonal persistence techniques & ideas in Houyhhnum Computing + +- Programming models (like better/safer primitives in C++, Pyth, Rust) +- Compilers +- Runtime Systems +- Handling Debugging Better + +Working on async bulk-synchronous actor-model parallelism + +1 Interested in graph algo, distributed machine learning + - Rust/C++ actor-based programming model +2 Pyth kernels using Intrepydd optimizing compiler + - Sparse matrix/tensor computation, pandas +3 Statistic & Dynamic Program Analysis Techniques + +# Florian Schaefer +## Inference and Computation|Competitive Optimization +- Inf & Comp +Conditional (normal?) distributions have a very deep connection to +factorizing matrices with Cholesky. +Creates fast algorithms for stats & physical simulations + +- Competitive Optimization +Sort of game theory, adversarial learning, and other. +Gradient Descent? Competitive Gradient Descent + - Machine Learning for PDEs + +Very cool! Very mathy!!! + +Interested in: +- Strong math affinity +- Programming experience with lots of new frameworks +- Likely 1-on-1 meetings +- Expects substantial effort on weekly basis +- Wants you to become published in a good venue (~1-1.5yrs) + +# Spencer Bryngelson +## Fluid dynamics for humans + +Algorithms for flows (mechanics/fluid dynamics) +- Computational Methods + +He works in the computational physics/supercomputing research group. +Industrial-scale simulations!!! (Like exascale/perfect scaling on a big +computer) + +Wants: +- Intermediate expertise in C/C++ or Fortran +- Strong handle on numerical linear algebra +- Knowledge of numerics for PDEs +Interested in: +- Computing at huuuge scale +- Using latest HPC hardware +- Solving pressing problems in health/defense + +# Thad Starner +## Contextual Computing Group + +Works on Google Glass woah. +Product originator, still works on new stuff. +Head Worn Display. + +- One research in sign language recognition stuff + - Significant problem for language abilities of deaf children + - Using MediaPipe for widescale sign-language recognition on a + smartphone +- Cetacean Hearing & Telemetry +- Research Group also studies: brain-neural interfacing + +Wants: +- Sophomores & Freshmen for long-term study +- Send email with UROC + Name + Project Interest to thad@gatech + +Not really my thing. + +Running List +------------ +- Hyesoon Kim (GPU/EE stuff) +- Rogues Gallery (but only the research side pls) +- Wei Xu (NLP) +- Anirudh Rama (community mesh network topology) +- Vivek Sarkar (Habanero) - somewhat cool but outside my wheelhouse +- Florian Schaefer (mathy guy) - yes yes yes +- Spencer Bryngelson (fluid dynamics & hpc) + +Possibly +- Niti's professor +- any other HPC researchers? |