I build production systems end to end, from retrieval architecture through to billing.
The lesson that shaped how I work: the first version of Vazmora's chat answered fluently, without proof. I rebuilt the retrieval pipeline so every answer traces to an exact source page, and a relevance threshold returns not found
rather than guessing. That instinct, to measure first and never let a model bluff, runs through everything else I build.
At PorsAI I lead the team building an enterprise AI orchestration platform: agents, voice, chat, workflows and knowledge on one governed runtime. I'm drawn to agentic systems, retrieval grounding and inference efficiency, and to the unglamorous parts in between: tenant isolation, latency regressions, eval loops on real transcripts.
e(avinash) ∈ ℝ64. Deterministic, decorative, hover a cell.
model_card.mdv2026.09
- name
- Avinash Vikram Singh
- architecture
- AI/ML engineer · founder · full-stack
- currently
- Founding Engineer @ PorsAI, leading the team behind its enterprise AI orchestration platform
Founder @ Vazmora
- focus
- agentic systems, retrieval grounding, inference efficiency
- training
- MSc Data Science, Christ University
BSc (Hons) CS, Delhi University
- awards
- Reliance Foundation Postgraduate Scholarship, 2025. Selected from a competitive national pool.
- context
- Bengaluru, India · IST
- known limitations
- Will rebuild your retrieval pipeline if it answers without a citation.