aabid mohamed a · backend · distributed systems & ai infrastructure
I build backend systems and AI infrastructure.
Distributed systems, developer tooling, and production-grade AI applications.
Python · Go · Distributed Systems · AI Infrastructure
Flagship repos are the fastest path to judge engineering depth — source, tests, ADRs, and case studies on this site stay aligned via shared catalog data.
Voice AI stacks split across transports and lack production operability (eval, replay, deploy).
Open-source voice AI infrastructure platform with one unified pipeline, RAG, evaluation, replay, and one-command HTTPS deployment.
Solo design and implementation — architecture through HTTPS deploy, tests, and docs.
One VoicePipelineService for WebSocket, REST, and LiveKit without duplicating agent and RAG logic.
Bug reports lack evidence and teams cannot audit how recordings become AI-written reports.
Local-first, AI-native bug reporting platform — screen recording upload, ffmpeg frame extraction, AI-drafted reports, shareable links, and an ordered Kafka pipeline in Go.
End-to-end Go systems build — API, workers, Kafka pipeline, observability, ADRs.
Ordered, idempotent media + AI pipeline per recording under Kafka replay and failure.
Unstructured forum and issue threads hide product and market signals; manual synthesis is not reproducible.
Source-agnostic pipeline turning GitHub and Reddit discussions into schema-validated market intelligence.
Pipeline design, collectors, normalizer, schema-bound extraction, evaluation harness, and FastAPI service.
Reliable LLM extraction with JSON Schema contracts and golden-dataset regression scoring.
Dealers need sub-minute wall previews and lead capture on unreliable shop-floor networks.
Full-stack paint dealer platform — upload a room photo, preview 63 catalog shades on walls in under 60 seconds, capture leads, and run quotes, orders, inventory, and receivables offline-first.
Full-stack product build — offline-first UX, Express API, ledger integrity, deploy.
Offline-first CRM sync with server-authoritative financial data.
Teams need a scheduler that survives worker crashes without double execution or lost jobs.
Upcoming distributed job scheduler and execution engine in Go — Postgres-backed leasing, fencing, and observability.
Planned — distributed job engine in Go (design phase; not yet implemented).
Postgres as source of truth with leasing, fencing tokens, heartbeats, and fair retry/DLQ semantics.
associate software engineer · backend + ai
building an agentic deal platform for an enterprise banking client — async ingestion, nlp extraction, ml classification, sales-desk apis, per-stage evaluation logging, and azure deploy discipline (client details confidential).
founder / principal engineer
solo-built open-source voice ai infra — unified ws/rest/webrtc pipeline, langgraph agents, pgvector rag, eval + replay, and one-command https deploy. metrics pinned in voxforge docs/project-metrics.md.
B.Tech · Electrical and Electronics Engineering
Long-form notes grounded in repos I built — not framework tutorials. Published articles will link here; until then, these topics match active implementation work on GitHub and case studies on this site.
in progress
Hiring for backend, distributed systems, or AI infrastructure? Email me with the system problem — I’ll point you at the repo, tests, and case study that best match it.
set in instrument serif for the masthead, ibm plex mono for everything else. rebuilt often. the field begins as deterministic static and animates on capable desktops. last updated November 2026.