CASE FILE 01
VOICECLAW
A voice agent built to increase tradesperson productivity: hands stay on the job while the agent handles the talking.
TRADES PRODUCTIVITY
HANDS-FREE
The work never stops for a screen.
Teaching machines to listen and speak.
I am Mahimai: a voice AI engineer who builds real-time agents, publishes the tools as open source, and writes about how it all works.
Voice-first AI systems that listen, understand, and respond. From real-time speech to production infrastructure.
Conversational AI that listens, understands, and responds. Real-time voice agents for customer support, sales, and operations. On-prem and self-hostable by design.
RAG systems, LLM orchestration, and document intelligence. From ingestion to production-ready answers.
Scalable architecture on Azure, AWS, and GCP. Built to handle voice traffic at scale, with observability baked in.
Things I have built and shipped: named products with real users, plus a live agent you can talk to right now.
A voice agent built to increase tradesperson productivity: hands stay on the job while the agent handles the talking.
The work never stops for a screen.
A Swiss apartment assistant serving residents 24/7: questions, requests, and escalations handled by voice, on infrastructure the client owns.
Residents get answers at 3am.
Do not take my word for it. Talk to a live voice agent right now and judge the latency, the interruptions, and the reasoning yourself.
The demo is the resume.
Seven roles, three arcs, every claim expandable. The full battle log from data science origins to running a voice AI agency.
CURRENT ARC
2025 - NOWMahimai AI Solutions
NA + EU
solo agency, clients retained 1yr+
Healify AI
1M REQ/MO
production LLM voice + chat APIs
Autobridge Systems
MTTR -60%
standardized logging + rollback
TRAINING ARC
2023 - 2025Future Smart AI
3 POC → CLIENTS
5 production demos shipped for GTM
Textify.AI
250K DOCS/MO
Azure Document Intelligence pipelines
ORIGIN ARC
2019 - 2022National University of Singapore
92% ACC
real-time CV safety detection on edge
S.Tech Solutions
-60% MANUAL CHECKS
ML on edge for manufacturing analytics
Code I have put into the world: repos people star, and packages you can install right now.
A curated map of notes, links, and patterns for people building voice AI agents.
292stars on GitHubCOMPUTER VISIONvechicle-counting-yoloCounts and measures vehicle traffic in real time with YOLOv5 and DeepSORT.
84stars on GitHubCURATIONawesome-csm-1bCurated use cases built on Sesame's CSM 1B speech model.
74stars on GitHubSEARCHlocallensSemantic file search engine for AI agents. 100% offline.
10stars on GitHubRUNTIMEopenrtc-runtimeRun multiple LiveKit voice agents in one shared Python worker process.
8stars on GitHubOBSERVABILITYvoicegatewayDrop-in observability and cost-control layer for LiveKit voice agents.
7stars on GitHub$ pip install voicegateway$ pip install openrtc$ pip install fastrtc-whisper-cpp$ pip install fastrtc_canary$ pip install vapiserve$ pip install locallens$ pip install envoicA book on production voice AI, and peer-reviewed research.
A RAG-based Medical Assistant Especially for Infectious Diseases
ICICT 2024, Lalitpur, Nepal, pp. 1128-1133READ PAPER ↗Classification of diabetes using ensemble machine learning techniques
Scalable Computing: Practice and Experience 25(4), 3172-3180READ PAPER ↗Freelance commissions: scoped, shipped, and closed. Six of the most recent.
Refactored an LLM pipeline with Map-Reduce patterns, cutting latency 50% and improving agent reliability.
Delivered a fully open source, on-prem voice assistant for local business automation.
Helped a startup ingest and index thousands of PDFs with a robust vector-store RAG system.
Built a Slack-integrated internal support assistant on Assistant v2, deployed with Docker.
Fixed critical dependency issues in an unmaintained research repo, unblocking academic deployment of privacy-preserving models.
Automated a production deployment pipeline with Kong API Gateway, AWS EKS, and a FastAPI backend.
Stats, abilities, and seals. The full spec sheet.
Voice is the primary interface for most viable exchange. My mission is to help people communicate better using Voice AI.
ALL SEALS LINK TO VERIFIABLE CREDENTIALS
Want to talk voice AI, open source, or anything else? My inbox is open.
Sharing knowledge is the only way technology grows.