Skip to content

Billel Sebaa

Process lead turning operational friction into working systems

Process lead at Infosys BPM with a computer-systems engineering background, focused on AI-assisted process improvement and digital transformation.

The person

About

As a Process Lead at Infosys BPM I work where operations meet technology: mapping how work actually flows, finding where it breaks, and rebuilding it so people can own their part of it without fighting the system.

My academic foundation is in computer systems engineering and management, which is why I tend to approach process problems as design problems. I am most useful when a workflow has accumulated friction nobody can quite name yet — order management, procurement, knowledge management — and it needs someone to take it apart, make the case for changing it, and then stay long enough to make the change stick.

I am currently focused on AI-assisted process improvement: not adding tools for their own sake, but removing the steps that only exist because software was inflexible.

Domain
Process improvementOrder managementProcurementKnowledge managementE-commerce operationsLogistics
Tech
TypeScriptReactPythonFirebase
AI
AI literacyPrompt engineeringLLM agentsRetrieval-augmented generation
Practice
Design thinkingTeam coachingStakeholder communicationEvaluation and guardrails
Languages
EnglishArabicFrenchPolish

Łódź, Poland

Career

Experience

  1. Senior Process Lead · Infosys BPM

    Leading process work end to end — diagnosing where a workflow loses time, making the case for the change, and staying until it holds without me.

    • Promoted through three process roles in under four years, each step on evidence of measurable improvement rather than tenure.
    • Own the process design decisions that other leads escalate — the ones where a fix in one team breaks a handoff in another.

    Free-form notes can go in this body. It is rendered after the highlights, so it is a good place for context that does not fit a bullet — constraints, team shape, what was inherited. Leave it empty and nothing is rendered.

    Łódź, PolandProcess improvementTeam coachingStakeholder communicationAI literacy

  2. Process Lead — Order Management · Infosys BPM

    Took ownership of order management as a lead, turning a role built on problem-solving into one built on prevention.

    • Transitioned from Process Specialist to Process Lead on demonstrated problem-solving and process optimisation, recognised by management.
    • Coached team members on process ownership, so exceptions were handled at source instead of escalating to a lead.
    • Built the analytical habit on the team — decisions argued from data rather than from escalation volume.

    Łódź, PolandOrder managementProcess improvementTeam coaching

  3. Process Specialist · Infosys BPM

    The two and a half years where the process knowledge that later became a lead role was built — order management exceptions, root causes, and the documentation that stopped them recurring.

    • Handled the exception paths that no standard procedure covered, then wrote the procedures that did.
    • Contributed to knowledge management so the team's tacit fixes became shared, findable process.

    Łódź, PolandOrder managementProcess improvementKnowledge management

  4. Buyer · ESPIR S.C.

    Owning purchasing decisions end to end — supplier contact, timing, and the cost of getting either wrong.

    • Moved from assistant to independent buyer in fifteen months across three roles.

    Białystok, PolandProcurementStakeholder communication

  5. Junior Buyer · ESPIR S.C.

    First role with real purchasing responsibility — supplier communication and order follow-through under supervision.

    • Built the supplier relationship habit that later made independent buying possible.

    Białystok, PolandProcurement

  6. Assistant Buyer · ESPIR S.C.

    Supporting the purchasing function — the admin layer where you learn how a supply chain actually behaves.

    Białystok, PolandProcurement

Education

  • MSc, Management — Smart & Innovative Business

    Politechnika Białostocka · 2020 · Białystok, Poland

  • MSc, Computer Systems Engineering

    University of Batna 2 · 2019 · Batna, Algeria

  • BSc, Information Systems and Software Engineering

    University of Batna 2 · 2016 · Batna, Algeria

Certifications

  • OpenAI Technical Practitioner · OpenAI
  • Design Thinking Practitioner
  • Supply Chain Management in Telecom Industry
  • Certificate of Achievement for Improvements and Enhancements in the Process
  • International Crossroads of ICT and Web Professionals in Algeria

Off the clock

Selected work

  • Vezaro

    2026 · In progress

    A cash-on-delivery store platform for Algerian merchants — storefront, back office, couriers and ad tracking in one system.

    Built for how commerce in Algeria actually runs: customers arrive from Facebook and TikTok ads on a phone, nearly every order is cash on delivery, and a lead only becomes a sale after someone calls to confirm it. Vezaro treats that as the real funnel rather than a checkout hand-wave — orders are sent to Facebook and TikTok as server-side conversions so ad budgets optimise on confirmed purchases instead of form fills, with junk-order protection, courier integrations for shipping and tracking, and a back office that splits confirmation work across a team with per-tab permissions. The storefront is trilingual by necessity (Arabic, French, English) with wilaya and commune level delivery pricing, so a merchant can launch a product page without a developer. The admin carries the rest: product setup, an AI page generator, orders and callbacks, delivery fees per wilaya, payouts per courier, and live ad spend against profit.

    Role · Founder — product, design and build

    TypeScriptReactFirebaseLLM agentsE-commerce operationsLogistics

  • Aegis — autonomous order entry

    2026 · In progress

    An agent that takes order cases from CRM through quoting and into SAP, reading the attached documents itself — and stopping for a human when money moves too far.

    Order entry is the kind of work that looks clerical and is actually judgement: read the PO, reconcile it against the quote, notice the logistics code was written in plain English, and push a payload the ERP will accept. Aegis does that chain end to end — parsing PDFs, Excel and email threads, correcting line items against the source document, and generating nested SAP S/4HANA payloads with the right partner functions and pricing conditions. The interesting part is what it refuses to do. Financial guardrails halt execution when a price correction exceeds thirty currency units or erodes margin by more than five percent, and hand the case back to Sales instead of guessing. Missing master data becomes a child case in the right queue rather than a silent failure, and the whole run is streamed to a live terminal so the operator can see exactly which decision was made and why.

    Role · Solo

    PythonLLM agentsOrder managementProcurementEvaluation and guardrails

  • Conversion Recover Agent

    2026 · In progress

    Analytics for store owners that explains why checkout traffic drops off, and proposes fixes with a forecast of the revenue they would recover.

    Store owners can see that people leave; they rarely see why. This tool instruments the checkout as a sequence rather than a page — scroll depth, time on form, where the flow stalls, which devices repeat — and turns those signals into named reasons for drop-off: a delivery fee that only appears at the last step, a form that is one field too long, a confirmation that never reassures. Output is deliberately prescriptive: prioritised changes to copy, layout or flow, each with an estimated revenue recovery rather than a vanity chart. It also flags spam and duplicate orders, which on cash-on-delivery funnels quietly corrupt every conversion number a merchant makes decisions from.

    Role · Solo

    TypeScriptLLM agentsE-commerce operations

  • SOP Assistant

    2026 · In progress

    An assistant that answers only from a team's written procedures, cites the section it used, and says "I don't know" instead of guessing.

    Most internal assistants fail in the same way: they sound certain when the procedure says nothing about the case. This one is built the other way round. Answers come only from retrieved SOP sections and carry a citation; if retrieval finds nothing relevant, it refuses rather than improvises. Input and output guardrails block off-topic questions, prompt injection and unsupported claims, and it will not advise skipping an approval or a credit hold. Every change is measured against an eval set before it ships — the current run scores 96% correct on 50 real-style questions and 100% on 18 adversarial ones, with zero leaks and zero false blocks. It is portable by configuration: one deployment per company, given a profile, its SOPs and its own eval set, with no real company documents in the repository.

    Role · Solo — design, build and evaluation

    PythonRetrieval-augmented generationLLM agentsEvaluation and guardrailsKnowledge management

Say hello

Contact

Tell me what is not working yet, and what you have already tried. If you are running a store and losing orders somewhere between the ad and the confirmation call, say so — that is the problem I work on.