
2
1
India
2+ years of experience
Mohamed Ahil Abdul AI/ML Engineer passionate about building practical AI solutions that solve real-world problems. I enjoy working with machine learning, computer vision, generative AI, and intelligent automation from idea to working product.

Warehouse receiving still runs on scanners, clipboards and memory. Workers have their hands full, so counts get typed in later, damage goes unreported, and nobody tells the supplier what was missing, so the company pays for goods it never received. DockVoice lets the worker just talk. They say the PO number and what arrived, in natural speech: "eighty controllers, sorry, eight zero... three of the sensors are cracked." The agent opens the purchase order, records the counts (a correction replaces the old number), splits damaged units into quarantine, and reads everything back. Only after a spoken "yes" does it post the goods receipt, update inventory, log shortages and damage, and email the vendor with a replacement request and due date. Safety is built in: the agent edits a draft, the post tool is only revealed after the read-back, any change after "yes" requires a new read-back, and every tool call is recorded in an audit log. Built on the AssemblyAI Voice Agent API: Universal-3 Pro streaming speech-to-text, turn detection and barge-in, tool calling, keyterms fed from live ERP data, and far-field noise isolation for loud docks. Stack: React + TypeScript, FastAPI, PostgreSQL (Supabase), deployed on Vercel and Render.
30 Sep 2026