// about

The person behind
the code

Shehryar Sohail speaking at a conference

I'm Shehryar. Most people call me Sherry. I'm a senior AI engineer from Pakistan who ships production LLM and voice AI systems.

Most recently I built a multi-agent voice platform for an insurance company that answers 10% of its live inbound calls at sub-800 ms average latency, with under 10% handed to a human. Before that I co-founded VECTOR Inc. as CTO, building LLM, speech and computer vision systems for clients.

I care about the gap between AI demos and AI that holds up in production. Most of the work is the boring part: evaluation, latency budgets, monitoring, and knowing when to hand off to a human.

On the side I write open-source Python and TypeScript tools, with 26,000+ combined downloads, and build products like WebLM. My resume has the full picture.

// Stack

Languages

PythonTypeScriptJavaScriptSQLC++

LLMs & Agents

RAGMulti-agent OrchestrationTool CallingStructured OutputLangChainLlamaIndex

Voice AI

ElevenLabsOpenAI RealtimeFaster-WhisperSilero VADSIP (PJSIP)3CX

ML & Fine-tuning

PyTorchHugging FaceLoRA/PEFTOpenAI Fine-tuning

Data

PostgreSQLMongoDBMilvusChromaDBPinecone

Backend & Cloud

FastAPINext.jsDockerAWSGCPCI/CD

// Journey

Oct 2024 - Oct 2026

Senior Python AI/ML Engineer, AI Team Lead · Paklogics

Built a multi-agent voice AI system for an insurance company: an orchestrator routing live callers to 4 specialist agents, handling 10% of inbound calls at sub-800 ms average latency with under 10% human handoff. Led a 3-engineer AI team and trained 5+ junior engineers and interns.

Jan 2023 - Aug 2024

Co-founder & CTO · VECTOR Inc.

Co-founded an AI services firm doing LLM, speech and computer vision work. Shipped SkyLocker, a WhatsApp assistant using multimodal RAG; fine-tuned LLMs (Falcon-7B, CodeGen-2B with LoRA, GPT-3.5/4o); built natural-language-to-SQL systems and Converso, a real-time voice bot. Won the DevFest 2023 pitch competition.

Jan 2022 - Jan 2023

AI/ML Research Trainee · AIMs Lab, FAST-NUCES

Fine-tuned BERT, XLM-RoBERTa, GPT-2, ResNet and ViT for text and image classification; contributed POS tagging and NER to an Urdu NLP library.

2020 - 2024

BS Artificial Intelligence · FAST-NUCES

Silver Medal for the second-highest semester GPA in the university; Dean's Honor List for 4 semesters.