Skip to content
View ANDROIDHASSAN's full-sized avatar

Block or report ANDROIDHASSAN

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ANDROIDHASSAN/README.md
Hassan Kazi. Voice AI Engineer: real-time voice agents and Indic speech.
Real-time voice agents, 1,000+ concurrent AI interviews, Indic speech, paper reimplementations

LinkedIn Portfolio Email LeetCode CodeChef

Followers

9x hackathon winner. 11 merged open-source fixes in LiveKit, Piper and Mastra. 1,000+ concurrent AI interviews in production. CodeChef 3 star.

About Me

I build real-time voice agents, and I'm chasing one question: why do voice agents break on real Indian speech, and how do we fix it?

  • 🏢 Now: Product Developer at InCruiter, leading the AI & real-time stack behind InCBot and IncScreen
  • 🏆 9× hackathon winner · CodeChef 3★ · before InCruiter: AI Engineer at myRide (Boston, remote)
  • ⚡ Scale: AI interviewers running 1,000+ concurrent interviews in 42+ languages
  • 🐛 Found in prod: English-only guards scored Devanagari 0.39 vs a 0.5 threshold; Marathi STT hallucinated in 10 of 15 turns
  • 🔧 Open source: 11 merged fixes in LiveKit (agents, python-sdks), Piper TTS, Mastra and KoljaB's RealtimeSTT / RealtimeTTS / stream2sentence (Indic and Urdu text); all merged PRs
  • 🔭 Building: my own Hindi/Marathi STT + TTS + LLM and a from-scratch streaming orchestrator
  • 📄 Learning: rebuilding transformers, Whisper, VITS and Moshi from the original papers
  • 📫 Reach me: hassankazinasik@gmail.com
Animated voice agent orb: listening and speaking

🖥️ hassan@voice-lab

hassan@voice-lab: Voice AI Engineer at InCruiter; 1,000+ concurrent interviews in 42+ languages; LiveKit Agents, Vapi, WebRTC; building own STT, TTS and LLM.

How a Production Voice Agent Works

Architecture of a real-time voice interviewer: candidate over WebRTC to a LiveKit room; Python agent worker with VAD, Sarvam STT, GPT-4.1 with tools and streamed TTS; script-aware guards and feature flags; CI/CD to LiveKit Cloud.

Building in Public

own-voice-models: my own STT, LLM and TTS with a streaming orchestrator. Building. indic-guard: script-aware text matching for Devanagari and romanized replies. Next.
IndicVoiceBench: open benchmark of STT and TTS on Indian phone audio. Next. Hinglish Turn Detector: end-of-turn model for code-mixed speech. Planned.
Voice agent pipeline with the project that targets each stage and its status. Target voice-to-voice latency budget of about 800 ms, not yet measured.

📄 Rebuilding the AI Stack from the Papers Up

🧠 LLM 👂 ASR 🗣️ TTS 🔊 Audio
Transformer · BPE Whisper VITS + HiFi-GAN SoundStream (RVQ)
LLM.int8 · GPTQ · FlashAttention Conformer + RNN-T StyleTTS 2 Moshi
LoRA · DPO · GRPO · speculative decoding wav2vec 2.0 F5-TTS Voice Activity Projection

One repo per paper, each with notes on what I learned.

Toolbelt & Tech Stack

🎙️ Voice AI

LiveKit Agents WebRTC Vapi Sarvam AI Azure OpenAI ElevenLabs Deepgram Silero VAD

🧠 ML & LLMs

PyTorch Hugging Face ONNX Whisper LoRA Claude

⚙️ Engineering



🕒 Recently Pushed

  • post-training — LoRA, DPO and GRPO implemented from scratch on a hand-written Llama/Qwen2… · 2026-10-14
  • speculative-decoding — Speculative sampling (Leviathan et al. 2023) from scratch on a hand-written… · 2026-10-14
  • quant-flash — LLM.int8(), GPTQ and FlashAttention implemented from scratch in PyTorch (+ a… · 2026-10-14
  • whisper-from-scratch — Whisper reimplemented from scratch in PyTorch: exact audio front-end, encoder-… · 2026-10-14
  • attention-is-all-you-need-hindi — Transformer (Vaswani et al. 2017) from scratch in PyTorch: English to Hindi… · 2026-10-14

✍️ Writing & Open Source

  • 📝 Why voice agents break on Hindi: 4 failure modes (draft)
  • ✅ 11 merged fixes (all):
    • LiveKit · agents #7741 recorder no longer hangs when the encoder fails, #7742 XML depth check · python-sdks #860 webhook sha256 claim, #861 Twirp base path behind proxies, #862 AudioMixer survives a late frame
    • Piper TTS (Open Home Foundation) #328 Hebrew phonemizer in inference
    • Mastra #27102 Sarvam listen() uses the configured model
    • RealtimeSTT #329 final full stop for Hindi/Bengali/Tamil, #330 feed_audio on tiny chunks · RealtimeTTS #383 ElevenLabs voice settings · stream2sentence #16 Urdu sentence splitting
  • 🔧 In review: LiveKit, Pipecat, NVIDIA NeMo, sherpa-onnx, Piper, Sarvam skills (open PRs)
  • 🛠️ Earlier I shipped CRMs: Interify · CupCount

GitHub Analytics

GitHub streak

Contribution snake

🎙️ From Hassan Kazi — let's build voice AI that understands India.

Popular repositories Loading

  1. Skillshare Skillshare Public

    HTML 3 1

  2. Meme-generator Meme-generator Public

    I have made a meme generator which will create random image memes just by one word

    Python 1

  3. awesome-chatgpt-prompts awesome-chatgpt-prompts Public

    Forked from f/prompts.chat

    This repo includes ChatGPT prompt curation to use ChatGPT and other LLM tools better.

    JavaScript 1

  4. code-for-freedom code-for-freedom Public

    A collaborative Independence Day project built by Consistent Coders, featuring two versions : one in React and one in HTML, CSS & JS . where everyone can contribute in their preferred tech stack. L…

    JavaScript 1 2

  5. The-Consistent-Coders The-Consistent-Coders Public

    This is a inituative for beginners , freshers and freelancers to crack their first internship / job offer by building actual projects with real users

    JavaScript 1 2

  6. python-sdks python-sdks Public

    Forked from livekit/python-sdks

    LiveKit real-time and server SDKs for Python

    Python 1