ammar.sheikh
All case studies
Active DevelopmentApplied AI · Retrieval2025 — present

Aether

Retrieval-grounded reasoning over private corpora, with citations strong enough to be auditable.

Role · Lead engineer

Overview

A retrieval-grounded reasoning system for organizations that need answers from their own documents — with citations strong enough to survive audit.

Problem

Off-the-shelf chat is a poor fit for enterprise corpora: stale contexts, hallucinated citations, and no way to measure regressions when prompts or models change.

Solution

A retrieval pipeline with versioned indexes, structured citations, and a continuous evaluation harness that gates deploys on retrieval quality and answer faithfulness — not just vibe checks.

Tech stack

  • TypeScript
  • Python
  • PostgreSQL
  • pgvector
  • LLM tooling

Engineering decisions

  • Evaluation harness is the unit of progress; nothing ships without a measurable retrieval or faithfulness delta.
  • Citations are first-class — produced by the model, verified against the retrieved set, and shown inline.
Engineering notes

What's shaping this build.

Design goals, philosophy, planned architecture, and where the project stands today. No fabricated benchmarks — only what's actually driving decisions.

Design goals
  • Citations that are verifiable, not decorative.
  • Evaluation that catches regressions before deploy.
  • Reproducible indexes so answers are attributable to a corpus version.
Philosophy

If you can't measure it, don't ship it. Retrieval quality and answer faithfulness both need offline scores that gate releases.

Planned architecture

Hybrid sparse+dense retrieval over chunked, versioned documents in pgvector; a reasoning layer producing chunk-level citations; an offline eval harness scored against curated question sets.

Current stage

Retrieval pipeline and eval harness scaffolded; iterating on chunking strategy and citation verification.

Links

Repository and demo links will appear here once the project is ready for public review.