The Austin Python Meetup

New location at Brazos Place @ 800 Brazos st. us the entrance on brazos street not 8th.
This is a Hybrid event both in-person and virtual. The in-person event is hosted at Antler downtown Austin.

Address:
Brazos Place @ 800 Brazos st,
The entrance is off of Brazos st.

Talks: Zombie Radio AI - Theatrical improvisation by LLM actors and Human quality Text-to-Speech for a Halloween party

By: Alfredo Valles

Description: Abstract:

Zombie-Radio: An interactive, audio-only theater play performed live by 4 AI voice actors: scientists trapped in a lab during a zombie breakout, broadcasting their struggle over the radio, a throwback to Orson Welles' 1938 War of the Worlds. The twist: the radio occasionally listens. Talk to it (push-to-talk), and the characters answer back, remember your name, and ask for your help.
Three AI models work in concert: an LLM (Nemotron Nano, served by llama.cpp) improvises the dialogue, a voice-cloning engine (Qwen3-TTS) speaks it in four distinct human voices, with a wide range of emotions, and Whisper listens to the audience so they can talk back to the cast in real time. A director (pure Python) conducts the story: it decides who speaks, when events strike and how the mood drifts, constrains the model's output with a grammar so every line arrives in screenplay form, streams lines to the voice as they are written, and trims the script so a show can run for hours without losing its thread. The TTS voices carry emotion, using reference audios from the EARS dataset, supporting up to 22 emotions.
This is a local first project: no AI APIs, and no pre-recorded audio. All AI inference runs on open-source models on a single GPU (Nvidia card +16GB RAM). Don't have an Nvidia card? No problem! We also provide Ansible installation scripts to quickly deploy our AI backend to any cloud VM that attaches an Nvidia GPU and runs Docker containers (e.g., Hyperstack).
The code — two open repositories:

• github.com/alfre2v/TalkWithZombies — the app: a fork of the open-source TalkWithMe that adds the show engine (the Python director, the grammar, the streaming voice, the listener's turn) and the browser stage, styled as an old radio set.
• github.com/alfre2v/zombie-radio — the infrastructure and the record: Ansible and Docker playbooks that stand up the three model servers on a GPU and install the app on a laptop, the tools that cast the voices, and the specs and design discussions from the whole build.

Come for the Halloween horror; stay for the engineering: grammar-constrained generation, a Python director that keeps an improvising model on track, emotional voice cloning, reproducible runs down to the byte, and what it was like to build it all with an AI pair programmer.

Bio:
Alfredo Valles is a backend engineer with years of experience. I love Python, Linux and open source. Alfredo is also a Physicist. Way back he used machine learning in research (before it was cool). Now he is dipping his toes in the brave new world of Transformers AI models.

Antler
Antler works with founders through our signature founder residency, adding value immediately, and taking no equity up front. We receive over 160,000 applications each year to build companies with us, and have grown to 18 funds with residencies in 30 cities around the globe. Founders also have the opportunity to pitch our Investment Committee to receive an up to $500K investment, allowing them to build for 12-18 months rather than attempting to raise before they have real traction.

New location at Brazos Place @ 800 Brazos st. us the entrance on brazos street not 8th.
This is a Hybrid event both in-person and virtual. The in-person event is hosted at Antler downtown Austin.

Address:
Brazos Place @ 800 Brazos st,
The entrance is off of Brazos st.

Talks: TBD

Antler
Antler works with founders through our signature founder residency, adding value immediately, and taking no equity up front. We receive over 160,000 applications each year to build companies with us, and have grown to 18 funds with residencies in 30 cities around the globe. Founders also have the opportunity to pitch our Investment Committee to receive an up to $500K investment, allowing them to build for 12-18 months rather than attempting to raise before they have real traction.

New location at Brazos Place @ 800 Brazos st. us the entrance on brazos street not 8th.
This is a Hybrid event both in-person and virtual. The in-person event is hosted at Antler downtown Austin.

Address:
Brazos Place @ 800 Brazos st,
The entrance is off of Brazos st.

Talk title: Manu Nicholas -Your Quantization Format Is Not Free: Measuring What On-Device LLMs Actually Cost

Talk description: When you download a quantized model you pick it by its filename. Q4_K_M, Q3_K_M, IQ4_XS. That name reads like a specification, but it is closer to a request. I built eight GGUF files from a single FP16 source and looked inside each one, and only three contained the format their name claims. Q3_K_M contained no Q3_K blocks at all. Q4_K_M at 0.5B parameters turned out to be 12 percent Q4_K by tensor count, storing 5.52 bits per weight against a nominal 4.5.
The gap is not academic. Two files carrying the same label decode up to 38 percent apart on the same machine, and the effect is large on a weak ARM core while nearly vanishing on x86, which is why it goes unnoticed on a developer workstation and then surprises people on a Raspberry Pi or a phone. This talk is about measuring that yourself. I will walk through the Python tooling I use to inspect a model file, show the numbers across three microarchitectures, and cover which formats sit off the efficiency envelope on every core I have tested. Everything is open and pip-installable, and reproducible on hardware you already own. Bring a model you actually use and we can look at what is really inside it.

Bio: Manu Nicholas Jacob is a hardware engineer at Dell in Austin, where he works in root-cause engineering on enterprise AI servers, looking at why systems fail and what the data actually says about it. He studied electrical and computer engineering at UMass Amherst.
Outside work he runs reproducible measurement studies on edge AI hardware, mostly Raspberry Pi 5 boards and laptops, on what really governs latency and energy: memory bandwidth, thermal limits, and quantization format. He maintains two open-source Python packages, llama-roofline and ml-systems-lab, reviews for the Journal of Open Source Software, and serves on the USENIX ATC 2026 artifact evaluation committee. He has a habit of publishing results that contradict his own earlier ones.
Happy to adjust length or depth for the room. Let me know what slot you have in mind and I will build to it.

Antler
Antler works with founders through our signature founder residency, adding value immediately, and taking no equity up front. We receive over 160,000 applications each year to build companies with us, and have grown to 18 funds with residencies in 30 cities around the globe. Founders also have the opportunity to pitch our Investment Committee to receive an up to $500K investment, allowing them to build for 12-18 months rather than attempting to raise before they have real traction.

New location at Brazos Place @ 800 Brazos st. us the entrance on brazos street not 8th.
This is a Hybrid event both in-person and virtual. The in-person event is hosted at Antler downtown Austin.

Address:
Brazos Place @ 800 Brazos st,
The entrance is off of Brazos st.

Talks: TBD

Antler
Antler works with founders through our signature founder residency, adding value immediately, and taking no equity up front. We receive over 160,000 applications each year to build companies with us, and have grown to 18 funds with residencies in 30 cities around the globe. Founders also have the opportunity to pitch our Investment Committee to receive an up to $500K investment, allowing them to build for 12-18 months rather than attempting to raise before they have real traction.