cv
General Information
| Name | Fiza Husain |
| Location | Bangalore, India |
| Focus | Speech AI, applied machine learning, model serving and evaluation |
Experience
- Founding Machine Learning Engineer Stimuler2025 — present
- Own the production speech stack end to end — automatic speech recognition, text-to-speech, and speech-to-speech — for a conversational English-fluency platform serving learners across India, Indonesia, and Latin America.
- First author of the INTERSPEECH 2026 paper introducing entity- and disfluency-aware evaluation for accented conversational ASR, raising entity recall from 53–55% to 80–85% and filler recall from under 5% to 76–86%.
- Fine-tuned and deployed region-specific LoRA adapters on audio language models for non-native English speech, served through vLLM at sub-second P95 latency.
- Built and scaled a low-latency streaming text-to-speech service on Ray Serve and Anyscale, including multi-model GPU partitioning and autoscaling.
- Designed the evaluation systems behind the stack — LLM-as-judge pipelines validated against human labels, error taxonomies, and metric hierarchies for conversational quality.
- Research Fellow M365 Research, Microsoft2023 — 2025
- Published research on intelligent monitoring and cloud operations, including first-author work at FSE Industry 2026 and co-authored papers at FSE 2024 and ICSE-SEIP 2024.
- Developed foundation-model approaches to monitor configuration and incident prediction over large-scale operational telemetry.
- Built statistical frameworks used to identify faulty, redundant, and misconfigured monitors across production cloud services.
- Analyst Goldman Sachs2022 — 2023
- Optimised a real-time graph pricer with market data subscription capabilities, improving data-driven pricing behaviour.
- Worked on the Halo pricing model for rapid response to underlyer price changes in securitized derivatives.
- Undergraduate Researcher Machine Learning Lab, IIIT Hyderabad2020 — 2022
- Research student with Prof. Praveen Paruchuri and Prof. Sujit Gujar on privacy and generalization in deep reinforcement learning.
- Developed a Privacy-Aware Inverse Reinforcement Learning analysis framework, published at AAAI 2022.
Education
- B.Tech (Honours), Computer Science and Engineering International Institute of Information Technology, Hyderabad2018 — 2022
- GPA 8.8 / 10.0
Selected Publications
- Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR. INTERSPEECH 2026 · first author
- Attention Enhanced Entity Recommendation for Intelligent Monitoring in Cloud Systems. FSE Industry 2026 · first author
- X-lifecycle Learning for Cloud Incident Management using LLMs. FSE 2024
- Intelligent Monitoring Framework for Cloud Services. ICSE-SEIP 2024
- How Private Is Your RL Policy? An Inverse RL Based Analysis Framework. AAAI 2022
- Full list with links on the publications page.
Patents
- Monitor class recommendation framework. United States patent application 18/593,338, 2025
Anjaly Parayil, Ayush Choure, Chetan Bansal, Saravan Rajmohan, Pooja Srinivas, Fiza Husain
Technical Skills
-
Machine learning
- LLM and audio-model fine-tuning (LoRA, DPO)
- Speech AI — ASR, TTS, speech-to-speech
- Evaluation systems and LLM-as-judge
- Reinforcement learning
-
Serving and infrastructure
- Ray Serve, Anyscale, vLLM, SGLang, Modal
- Streaming and distributed inference
- Docker, Kubernetes, AWS, GCP, BigQuery
-
Frameworks
- PyTorch, Transformers, PEFT, Hugging Face
- Weights & Biases
-
Languages
- Python, C++, SQL