Nabeel Hussain

Impact in production

Now

Senior AI Engineer, Conovo Technologies (Aug 2023 – Present, remote)

Focus

Multi-agent systems, voice agents, RAG, LLM evaluation and grounding

Stack

Python, FastAPI, Django, LangGraph, LangChain, PostgreSQL, Redis

Infra

AWS, Azure, Docker, GitHub Actions

Hours

Full EU overlap, 4–6 hours with US Eastern

Open to

Relocation to the Gulf or Europe

• 10×faster query responses on RAG systems, from hybrid BM25 and dense retrieval with reranking Conovo, RAG
• 35%cut in LLM API costs via multi-provider routing and semantic caching Conovo
• 1M+records processed per pipeline run, across 20+ daily Airflow pipelines Conovo, data
• 250+data sources running in production, from AWS S3 to the web Conovo, data
• 70%less manual work with document AI: OCR and NLP over scanned legal and financial documents ArhamSoft

10 specialist agents

PersonifAI

Multi-agent procedure analysis for regulated banks: consulting-grade findings in minutes, shipped on-premise.

• Ten specialist agents audit each procedure for compliance, controls, risk and automation readiness.
• Produces audit-ready reports and BPMN process maps.
• An anti-hallucination layer verifies every regulatory citation live against 11 UK and US regulator sources.
• Provider-agnostic across 4 LLM providers, with 1,300+ automated tests.

• Multi-agent LLM
• Citation grounding
• 4 LLM providers
• BPMN 2.0
• On-premise
• 1,300+ tests

Read the case study askpersonifai.com (opens in a new tab)

10,000 agents per request

Global Agentic Market Exchange

The G.A.M.E.: a neutral exchange where AI agents hire other agents and get paid per call.

• Agents call each other through an HMAC-signed gateway, with per-call pricing.
• A double-entry ledger holds each fee in escrow and releases it on a good response; the agent keeps 95%, paid out through Stripe Connect.
• Built the entry and anti-fraud systems for a $10M prize challenge: batch registration of 10,000 agents per request, liveness and uniqueness checks, near-duplicate detection.
• Designed to scale to 1M+ agents.

• Django
• PostgreSQL
• Celery
• Redis
• Stripe Connect
• AWS

Read the case study agenticmarket.exchange (opens in a new tab)

ProofDial

Independent audits for voice agents: every failure comes back with its line, its recording and the control it breaks.

• Audits the calls an agent already took, and places automated test calls against production voice agents.
• Scores every call with Claude as LLM judge.
• Redacts personal data with Presidio before anything is stored.
• Maps 13 failure types to 38 named controls, on a multi-tenant backend.

• Pipecat
• Deepgram
• Cartesia
• Silero VAD
• Claude
• Presidio
• Supabase

Read the case study proofdial.com (opens in a new tab)

7 stages, requisition to hire

SharpSource ATS

An applicant tracking system built as AI: the system does the work, recruiters review and decide.

• Writes the job description, screens and ranks every applicant, and schedules the interviews.
• Must-have skills gate eligibility; a weighted fit score shows its reasoning and confidence.
• Drafts the offer and forecasts whether it will be accepted.
• Every AI call can be overridden, and critical actions land in an audit log nothing can edit.

• AI screening
• Fit scoring
• Interview scheduling
• Offer forecasting
• Audit log

Read the case study sharpsource.global (opens in a new tab)

+30% performance

Webcurve

Counterfeit seller intelligence: a fraud detection platform that tracks sellers of counterfeit products.

• Led a team of three building the platform.
• Scheduled the scraper DAGs with Apache Airflow.
• Automated scraper maintenance with a custom-trained LLM.
• Query optimisation gave a 30% performance gain.

• Qwen2.5
• Claude 3.5
• Apache Airflow
• Scrapy
• PostgreSQL
• AWS

Read the case study Admin portal (opens in a new tab)

4 steps, data room to Golden Record

LiquidDocs

Expert-verified AI due diligence: a messy data room becomes a signed Golden Record.

• AI agents extract, structure and classify every document at machine speed.
• A named professional verifies, corrects and signs off every finding.
• Every finding is scored for risk and linked to its source, with a full audit trail.
• The same engine runs AI transformation for firms, built on the Vault.

• Due diligence
• Golden Record
• Expert sign-off
• Audit trail
• The Vault

Read the case study liquiddocs.ai (opens in a new tab)

−40% inference time

21Swap

Deep-learning face transformation for images and video.

• Led development of the face transformation system.
• GPU acceleration cut inference time by 40%.

• Python
• TensorFlow
• OpenCV
• Flask
• AWS GPU

AssetCurve

An enterprise platform for IT assets and hardware.

• Tracks equipment costs and recertifications.
• Built on Django REST Framework, with Stripe payments, on AWS.

• Django
• DRF
• PostgreSQL
• AWS
• Nginx
• Stripe

24/7 website and WhatsApp

Balie

An AI receptionist for growing businesses: it answers customers around the clock and fills the diary.

• Answers customer questions on the website chat and on WhatsApp, in Dutch and English.
• Books appointments straight into Google Calendar or Outlook, with confirmations and reminders.
• Trained on the business’s own site, PDFs, Notion and Drive; live in five minutes with one line of code.
• Captures leads, and hands the conversation to a person when needed.

• Website chat
• WhatsApp Business API
• Knowledge base
• Appointment booking
• Universal inbox

Read the case study usebalie.nl (opens in a new tab)

3 roles: nurse, investor, employer

NX Innoverse

A healthcare-innovation platform built around nurses: pitch, fund and hire in one place.

• Nurses showcase and pitch their innovations, projects and career opportunities.
• Investors discover and back the ideas they believe in.
• Employers post jobs and hire nursing talent.
• Pitches stay private, with confidentiality and intellectual property protected.

• Pitches
• Investments
• Job board
• Career moves
• Role-based onboarding

Read the case study nx-innoverse.com (opens in a new tab)