Skip to content
Tanveer
  • FounderLegacyLift
  • StatusBuilding full-time

Tanveer.
Building LegacyLift.

An 18-year-old self-taught full-stack developer building LegacyLift full-time during a gap year.

Explore LegacyLift

Modernise legacy systems. Prove nothing broke.

LegacyLift turns decades-old code into modern, tested software. It learns every business rule, migrates one piece at a time with a person approving each change, and proves the new system behaves exactly like the old one.

Built for banks, government, and the systems that can’t go down.

  1. See what the old code actually does.

    LegacyLift reads the source, extracts the business rules buried inside, and maps how everything connects, so nothing important is lost in translation.

    ACCTINT

    • callsINTCALC
    • callsDATEUTIL
    • readsACCTREC

    Business rules found

    • Interest tiers by balance bandFinance
    • Senior bonus over age 65Finance
    • Dormant after 12 months inactiveOps

  2. Rewrite it, with a human in control.

    Modern code, generated one piece at a time and reviewed old versus new. A person approves every change before it ships. Nothing merges on trust.

    FEECALC.CBL

    every change · no exceptions

    Legacy · COBOL
    0000-MAIN.  MOVE 0 TO LK-FEE.  IF ACCT-TYPE = 'S'    GOBACK.  PERFORM 2000-FEE.
    Migrated · Python
    def main_0000(rec):  lk_fee = Decimal("0.00")  if rec.acct_type == "S":    return lk_fee  lk_fee = fee_2000(rec)

  3. Prove the new system still behaves.

    The rewrite is replayed against the real behaviour of the old system and compared field by field.

    Legacy

    Migrated

    VectorLegacy = MigratedΔ
    ACC001standard1,201.500.00
    ACC004senior tier5,006.250.00
    ACC005dormant790.000.00
    ACC007premium25,041.670.00
    ACC010closed0.000.00

  4. Proof, not promises.

    The result is sealed with a signed record your risk team can verify. Something you can hand to risk and audit that actually holds up.

    100.0%

    behavioural equivalence

    signed ed25519

    • Value
    • Sequence
    • Store
    • Determinism
    • Attested

    Independent, reproducible, and signed. Ready for risk and audit.

Replay can teach the migrator, or certify the result. Never both in the same lineage.

The Independence Rule. A validator that grades a rewrite against the same data that produced it is marking its own homework. LegacyLift keeps the two apart, so a passing score means something to your model-risk team and to an auditor.

It found the bug nobody would notice.

COBOL silently cuts off numbers that overflow their field. A modern rewrite doesn’t, so the new system quietly returns different numbers. On a real COBOL program, LegacyLift caught exactly that, before it shipped.

Bring your own key. Client source never touches LegacyLift infrastructure.

Read the case study

What I build.

Pollen Mesh

Shared threat intelligence, without sharing raw data.

A privacy-preserving, federated threat-intelligence system. Organisations detect coordinated attacks across company boundaries by sharing only matched, hashed attack signatures, gated behind two rounds of human approval. Built in a single day at Cambridge’s Collaborative Agent Hackathon, hosted by Flower Labs.

Built with Shritesh Jamulkar, Software Engineer at Booking.com.

Won an Honourable Mention, and demoed it at Flower Labs’ Flower Monthly.

Visit Pollen Mesh

  1. HukamConnect

    I architected and deployed a multi-tenant platform for Sikh Gurdwaras, pairing NFC hardware with a Firestore backend for real-time content distribution and engagement tracking.

    300+ active users2,000+ monthly visits

    hukamconnect.com

  2. ArgusAI

    A voice-driven warehouse goods-receipt agent. Matches spoken delivery descriptions to purchase orders and escalates only exceptions.

    Built at the {Tech: Europe} Agentic AI Hack, hosted by Conduct and Google DeepMind.

  3. Scoofy

    I owned the frontend architecture for this AI e-commerce platform, rebuilding it in Next.js and TypeScript, closing a six-month backlog in four, and shipping LLM-powered recommendations and real-time data flows.

    Head of Frontend Development · Feb–Oct 2025

Bringing people together.

Sikhs in Tech

Director of Hackathons · London

I’m organising the first Sikhs in Tech London hackathon.

Scale
100 developers
Format
One day
When
Early November 2026

Let’s
talk.

hello@tanveersingh.dev