Class Syllabus
- This is a seminar course, specific for the writing requirement.
- Attendance to department seminars, and assigned readings.
- The student is to write original reports and a project report.
Requirements
- Join the class slack channel csc-courses.slack.com, #csc405-271.
- Requirements are,
- Three reports and a final project.
- Attend all the talks (or let me know for an excused absence)
- In person attendance (or let me know for a excused zoom attendance)
- Of the three reports, one must be a talk.
- Read the assigned readings ahead of class discussion, and participate in the class discussion.
- Note: Reports are due in two weeks from the in-class discussion or seminar being reported.
- Report word lengths, and requirement to rewrite, to be discussed.
- All work must be completed by midnight, the last day of class.
Report
Writing Assignment: Write a 1,000-word technical essay on the assigned topic.
The essay should present a clear thesis or central argument, be logically organized,
support its claims with appropriate sources, and use clear, professional technical writing.
Submission: Please submit the paper by DM to me on the class
csc-couses.slack.com, the #csc405-271 channel.
Draft and Revision: A preliminary draft will be submitted for review and feedback.
Students will revise the draft and submit a final version incorporating the feedback received.
Final Project: A 2,000-word paper on a topic related to any of the talks.
The paper must extend or take a new direction from an idea in the chosen talk, supported by relevant references,
Must be submitted by one week before the end of class to allow for a revison.
Grading
- Grading on a 0–5 point scale.
- 1/6 each: attendance and each of the three reports.
- 1/3 final report.
Assigned readings and talks, with date of class discussion or talk occurence
August 24:
- Alan Turing Computing Machinery and Intelligence, (1950)
The Imitation Game, rebuttal to Lady Ada Lovelace.
August 31:
- Prof. Morris Alper: Tracing the Signs: Multimodal Foundation Models for Ancient Writing
Sept 7:
- Labor day: vacation
Sept 14:
- Collin Connors: How Attackers Exploit AI Systems
Sept 21:
- Panos Leledakis: AI Unlocked:
Practical Superpowers for the Next Generation of Innovators
Sept 28:
- Erica Newcome: Mastering AI for Computer Science Research
Oct 5:
- Whitfield Diffie and Martin Hellman, New Directions in Cryptography (1976)
The discovery of public key cryptography
Oct 19:
- Michael Scott Fisher: Data-driven tools to improve tropical cyclone forecasts
Oct 26:
- Satoshi Nakamoto, Bitcoin: A Peer-to-Peer Electronic Cash System (2008)
The invention of Bitcoin by the enigmatic inventor.
Additional resources for the Whitfield Diffie and Martin Hellman paper:
- R. L. Rivest, A. Shamir, L. Adleman, A Method for Obtaining Digital Signatures and Public-Key Cryptosystems,
Communications of the ACM, 21(2), 120-126, February 1978.
- Robert J. McEliece, A Public-Key Cryptosystem Based on Algebraic Coding Theory,
JPL Deep Space Network Progress Report 42-44, Jet Propulsion Laboratory, 1978, pp. 114-116.
- Taher ElGamal, A Public Key Cryptosystem and a Signature Scheme Based on Discrete Logarithms,
IEEE Transactions on Information Theory, 31(4), 469-472, July 1985.
- Victor S. Miller, Use of Elliptic Curves in Cryptography, CRYPTO '85.
- Neal Koblitz, Elliptic Curve Cryptosystems,
Mathematics of Computation, 48(177), 203-209, January 1987.
- O. Regev, On Lattices, Learning with Errors, Random Linear Codes, and Cryptography,
Proceedings of the 37th Annual ACM Symposium on Theory of Computing (STOC), 2005, pp. 84-93.
Not yet assigned:
- The Rise and Fall of Agent Civilizations,
Dwarkesh Podcast.
- Philip Daian, Steven Goldfeder, Tyler Kell, Yunqi Li, Xueyuan Zhao, Iddo Bentov, Lorenz Breidenbach, and Ari Juels,
Flash Boys 2.0: Frontrunning, Transaction Reordering, and Consensus Instability in Decentralized Exchanges,
arXiv 1904.05234, 2019.
- Ernst Schroedinger What is Life?
- Quantum computing
Tracing the Signs — Multimodal Foundation Models for Ancient Writing
Morris Alper, Ph.D.
Ungar Building, Room UB305 (3rd Floor), Coral Gables Campus
Monday 08/31/2026 at 5:05PM | Zoom: https://miami.zoom.us/j/95533396867
Refreshments at 4:30PM
Abstract
Modern vision-language models (VLMs) can read handwriting, parse complex documents, and
translate text across dozens of languages. But when applied to photographs of ancient i
nscriptions, whether carved, incised, or impressed into clay, metal, and rock thousands
of years ago, they fail. Our ongoing research studies how foundation VLMs can be
adapted out-of-domain to handle complex ancient writing systems, with applications
to cuneiform, ancient North Arabian inscriptions, and potentially many others.
I will show how we can adapt diffusion models for pose estimation to estimate the
internal configurations of signs for paleography and improved optical character
recognition, efforts to create new benchmarks, and our vision for future models
bridging visual knowledge with ancient languages and epigraphy.
Bio
Morris Alper is an Assistant Professor in Computer Science at the University of Miami.
His research focuses on multimodal learning and AI applications for linguistics,
applied across textual, visual, and audio modalities. He joined the faculty this
month; previously, he was a postdoc at CMU's Language Technologies Institute,
after receiving his PhD in multimodal machine learning from Tel Aviv University.
How Attackers Exploit AI Systems
Collin Connors, PhD
Ungar Building, Room UB305 (3rd Floor), Coral Gables Campus
Monday 09/14/2026 at 5:05PM
Refreshments at 4:30PM
Abstract
Artificial Intelligence is transforming software development, cybersecurity, business operations, and
everyday life. As AI systems become more deeply integrated into applications, cloud platforms, and
critical infrastructure, they are also creating entirely new security challenges. From prompt injection
attacks and data leakage to autonomous AI agents performing unsafe actions, attackers are already
exploiting weaknesses in AI systems in ways that traditional security controls were never designed to
address. Using real world case studies we will explore how AI expands an organization's attack surface,
and why AI systems must be treated as both powerful tools and potential security liabilities.
Bio
Dr. Collin Connors is ERMProtect's AI Risk Management Lead. He guides organizations to implement
artificial intelligence strategically and securely, so they can maximize productivity and business
opportunities within compliant and best-practice governance frameworks. His work encompasses
enterprise-wide implementation of AI he helps organizations align their AI strategy to business goals,
establish forward-looking governance and risk management policies, and implement key IT controls to
prevent misuse and security incidents. Collin also developed ERMProtect's AI Audit program, conducts
risk assessments and AI-focused penetration testing. His AI clients span large financial institutions, city
governments, and academic institutions.
AI Unlocked: Practical Superpowers for the Next Generation of Innovators
Panos Leledakis, Founder @ IFAAcademy
Ungar Building, Room UB305 (3rd Floor), Coral Gables Campus
Monday 09/21/2026 at 5:05PM
Refreshments at 4:30PM
Abstract
A practical session on the AI capabilities most people
— including most computer science graduates — never discover,
and how to use them to build faster, work smarter, and stand out immediately in the job market.
-
Beyond the chatbot — what LLMs can actually do once you go past basic prompting:
agentic AI, persistent memory, tool use, and connecting models directly to real data and workflows.
-
Live demos of hidden features — capabilities inside Claude,
ChatGPT and other tools that almost nobody uses, shown live, not just described.
-
Building your own "second brain" — an external AI memory system,
giving you full LLM freedom, that knows your work, your projects and your goals,
and can guide you like your best consultant, advisor, or friend.
-
The software stack worth knowing — beyond the obvious tools:
what's free, what's underrated, and what's about to matter a lot more.
-
The career edge — how to use AI to build real, working projects fast,
stand out in interviews, and skip years of "figuring it out the hard way".
Live Q&A — bring your hardest AI questions.
Bio
Panos Leledakis has an extensive 27 years of experience in insurance and financial
planning since 1999, as a top-awarded insurance consultant and a TOT MDRT member.
A pioneer in the insurance industry and an insurance futurist, he is an expert in
digital transformation, artificial intelligence, internet marketing strategies,
automation, CRM, virtual and augmented reality, and cutting-edge technologies
in the insurance sector. He is the inventor of two innovative software solutions
for insurance needs analysis and financial planning, implementing the science of
risk management, artificial intelligence, and extensive neuroscience and
behavioral economics research on risk perception and decision-making.
His work has earned him 11 global innovation awards from organizations like MIT.
He holds a certification in "AI for Business" from the University of
Pennsylvania (Penn) and is furthering his expertise in artificial intelligence
through ongoing studies at MIT. He also holds certifications in LUTCF, IRMA,
BRMA, and is pursuing additional studies with PRMIA (Professional
Risk Management Association). In 2020, he was honored with the esteemed
global award for "Top 25 InsurTech CEOs of 2020."
A dynamic public speaker,
he has inspired audiences in over 45 countries with his motivational and
engaging style. Originally from Greece, he relocated to the USA, living in New York
and Miami, after 2020.
Mastering AI for Computer Science Research
Erica Newcome, M.S.
STEM & Interdisciplinary Research Librarian
Ungar Building, Room UB305 (3rd Floor), Coral Gables Campus
Monday 09/28/2026 at 5:05PM
Refreshments at 4:30PM
Abstract
This session will focus on best practices for integrating AI into research.
Participants will learn how to define research goals, choose appropriate AI tools,
prioritize ethics, and interpret results.
Bio
Erica Newcome is the STEM and Interdisciplinary Research Librarian at the University of Miami Libraries.
She has a Master of Library and Information Science from San Jose State University and a
Bachelor of Science in Conservation and Resources Studies from the University of California, Berkeley.
Prior to arriving at the University of Miami Libraries, she worked for the Earth Sciences & Map Library at UC Berkeley.
HMs. Newcome is the liaison to the College of Engineering along with the departments of Chemistry,
Computer Science, Mathematics, and Physics. She is dedicated to supporting academic
research by providing consultative and instructional support for students, faculty, and researchers.
A Virtual Hurricane Hunter: Learning 3D Hurricane Winds from Satellites and Sparse, Biased Labels
Michael Fischer, Ph.D.
Ungar Building, Room UB305 (3rd Floor), Coral Gables Campus
Monday 10/19/2026 at 5:05PM
Refreshments at 4:30PM
Abstract
Hurricanes can strengthen from a tropical storm to a Category 5 in about a day, and these
rapid-intensification events are where forecasts fail most.
The storm's three-dimensional wind structure holds clues to what comes next,
but the instrument that measures it, Doppler radar on hurricane-hunter aircraft,
has sampled only about 4% of the world's tropical storms since 1997.
Geostationary satellites image every storm every 30 minutes, but they see cloud tops, not winds.
This talk presents TC-SWARM, a machine-learning system that reconstructs a hurricane's
3D wind field from satellite infrared imagery and environmental data,
learning from an archive of 1,640 airborne radar analyses of 94 storms.
The labels make this an unusual learning problem: few independent samples,
targets observed over only about a fifth of a typical storm,
observations missing not at random, and the most damaging winds almost absent.
I will show how these constraints shaped the design: per-point regression with
gradient-boosted trees, a hierarchy of physically meaningful intermediate targets,
quantile models and stacking to counter regression to the mean,
validation on whole storms and an entire held-out season, and scores that
reward getting the structure right. On the 2025 season, the model halves the squared
error of standard vortex models given the same inputs, and Category 5 Hurricane
Melissa shows both where it works and where it fails. I will close with lessons
for machine learning on scientific data and a vision for a global, vortex-scale hurricane
reanalysis, including open problems where computer science can help.
Bio
Michael S. Fischer is an Assistant Professor at the University of Miami,
with a joint appointment in the Rosenstiel School's Department of Atmospheric
Sciences and the Frost Institute for Data Science and Computing (IDSC), where he is a core
faculty member. His research focuses on tropical cyclone structure and intensity change,
particularly the inner-core processes that govern rapid intensification. His work draws
on aircraft reconnaissance and satellite observations as well as machine learning,
with the aim of developing methods that transfer to operations. He built TC-RADAR,
a database of more than 1,600 radar analyses from NOAA's hurricane hunter aircraft,
and TC-ATLAS, an open-source web platform for exploring hurricane structure and intensity change.
He holds a Ph.D. in atmospheric science from the University at Albany,
State University of New York, and previously held research positions at NOAA's Hurricane
Research Division and the University of Miami's Cooperative Institute for Marine and Atmospheric Studies (CIMAS).
