# Akshat Kushwaha

> building things that didn't get permission to exist. AI systems engineer. Co-founder of bits&bytes, Advisory Board Member at Boardy, former Jr. Research Engineer at jhana, and previously founded Lexana.

Last updated: 2026-10-27

Note: blog posts on this site are drafted with an LLM and reviewed and edited by Akshat before publishing.

## Overview
Akshat Kushwaha (also known as a3ro, aero, Akshat Singh Kushwaha) is an Indian AI systems engineer. He trains models, builds retrieval and vector search, and ships low-latency AI backends.

## Current Roles
- **Co-founder** of [bits&bytes](https://gobitsnbytes.org) (Registered Section 8 foundation and builder network in India).
- **Advisory Board Member** at [Boardy](https://boardy.ai) (Advising on Boardy Pro and product systems for builders).

## Technical Core Competencies
- **Large Language Models & LLMOps**: Audio evaluation pipelines, speech-model benchmarks, context window managers for long drafts, latency reduction (60%+ TTFT improvement), agentic orchestration.
- **Backend & Systems**: Rust, Async FastAPI, Python, Docker, Redis, PostgreSQL, WebSockets, real-time collaboration engines, circuit breakers, and semantic caching.
- **Retrieval & Vector Databases**: Embedded vector search (Rust, AVX2 SIMD, wgpu GPU compute, immutable WAL), hybrid vector + graph search (FAISS HNSW, Okapi BM25, NetworkX, RRF fusion k=60), SQLite WAL, time-aware indexing.
- **Frontend & Product**: Next.js (App Router), React 19, TypeScript, Tailwind CSS, Framer Motion.
- **Model Training**: transformer design, imitation learning and distillation, LoRA/QLoRA fine-tuning, 4-bit quantization, on-device inference, evaluation and calibration.

## Selected Production Builds & Systems
- [Qenlo Embedded Vector Database](https://github.com/a3ro-dev/qenlo): Local, embedded vector store in Rust (beta on crates.io, PyPI, npm). One canonical record store decides what exists; CPU, WGPU, USearch, and PyTorch indexes are derived and rebuildable. Exact search by default with user/time filters applied before ranking, every mode checked against an FP64 reference. Research paper: The Efficient Kernel Runs Slow.
- [Reflex](https://github.com/a3ro-dev/reflex-2048): 14.5M-parameter transformer that plays 2048 with one forward pass per move and no search. Board as 21 tokens (16 tiles, 4 afterstate move tokens, 1 value token), 8 pre-norm blocks, trained by imitating an expectimax teacher over 460M moves. 164,395 mean over 10,000 fresh games, 2048 tile in 96.6%.
- [HybridMind Retrieval Engine](https://github.com/a3ro-dev/hybridmind): Time-aware local hybrid retrieval engine combining FAISS HNSW, Okapi BM25s, NetworkX directed multigraph, and SQLite WAL with RRF (k=60) fusion and Python SDK.
- [Luna Companion](https://github.com/a3ro-dev/luna): AI-native menstrual cycle companion with NLP chat logging, adaptive exponential smoothing with population priors, Supermemory v4 persistent memory, and OpenUI cards (Next.js 16, Drizzle ORM, Neon Postgres, Vercel AI SDK).

## Quick answers
### who is Akshat Kushwaha?
Akshat Kushwaha (a3ro) is an AI systems engineer from Prayagraj, India. he trains models, builds retrieval and vector search, and ships the backends around them. he built Qenlo, an embedded vector database in Rust, and Reflex, a transformer that plays 2048 without search. he started building at twelve, co-founded bits&bytes, and sits on Boardy's advisory board.

### what is Reflex?
Reflex (reflex-2048) is a 14.5M-parameter transformer Akshat Kushwaha trained to play 2048 with one forward pass per move and no search. it learned by imitating an expectimax teacher across 460M of its moves. over 10,000 fresh games it averages 164,395 points and builds the 2048 tile 96.6% of the time. code, weights, and paper are public.

### what has Akshat Kushwaha built?
his main builds are Qenlo, an embedded vector database in Rust; Reflex, a transformer that plays 2048 without search; HybridMind, a local hybrid retrieval engine; and Luna, an AI menstrual cycle companion. at jhana he built the audio evaluation pipeline for Steno and a GPT-4o-mini harness that cut time-to-first-token by 60%+.

### what is Qenlo?
Qenlo is a local, embedded vector store written in Rust by Akshat Kushwaha. one canonical record store decides what exists, and the CPU, WGPU, USearch, and PyTorch indexes are derived from it. search is exact by default, with user and timestamp filters applied before ranking. it's in beta on crates.io, PyPI, and npm, with a paper: The Efficient Kernel Runs Slow.

### how do i contact Akshat Kushwaha?
email akshatsingh14372@outlook.com. you can book a call at calendly.com/a3ro-dev/let-s-talk, or find him as a3ro-dev on GitHub, @a3rodev on X, and Akshat Singh Kushwaha on LinkedIn. he's based in Prayagraj, Uttar Pradesh, India, so replies land on IST.

## Builds compared
| build | what it is | stack | status | key number |
| --- | --- | --- | --- | --- |
| Qenlo | embedded vector database | Rust, exact CPU scan, WGPU, USearch, PyTorch | beta | every search mode checked against an FP64 reference |
| Reflex | transformer that plays 2048 without search | PyTorch, 14.5M-param transformer, batched GPU env | released | 164,395 mean over 10,000 games, 96.6% reach 2048 |
| HybridMind | local hybrid retrieval engine | Python, FAISS HNSW, BM25s, NetworkX, SQLite WAL | wip | RRF fusion at k=60 over dense, lexical, and graph ranks |
| Luna | AI menstrual cycle companion | Next.js 16, Neon Postgres, Drizzle, Vercel AI SDK | active | adaptive smoothing with population priors, 2.5σ outlier clamp |

## Published Notes & Essays
- [why i built qenlo](https://a3ro.dev/blog/why-i-built-qenlo): it started with a pretty normal question. i was talking to a friend and asked him what vector database he uses. "qdrant," he said. okay, but
- [architecting motherboard: polymorphic iam, forensic digital signatures, and operations at teen-builder scale](https://a3ro.dev/blog/architecting-motherboard): when you run an independent builder network across a dozen Indian cities with hundreds of teenage hackers, off-the-shelf SaaS tools fall apa
- [why luna needs to exist: adaptive math, memory, and the problem with pink calendar apps](https://a3ro.dev/blog/why-luna-needs-to-exist): most cycle tracking apps on your phone are insulting. they fall into two categories, and both of them are built on deeply patronizing assump
- [architecting lexana at my nani's house: 4-agent swarms, 60% semantic cache hits, and sub-200ms p95 latency](https://a3ro.dev/blog/architecting-lexana-at-nanis-house): in the summer of 2025, i spent nearly two months at my nani's place in Prayagraj. there was no broadband connection. there was no ergonomic 
- [why we need hybridmind: vector search is blind, graphs are brittle, and AI memory needs local ground truth](https://a3ro.dev/blog/why-we-need-hybridmind): if you ask most engineers how they handle long-term memory for an AI agent, they will tell you: > "we embed the conversation history into Op
- [Inside an Autonomous Hackathon Screening Agent: Architecture, Scoring, and Failure Modes](https://a3ro.dev/blog/autonomous-ai-eval-cli): India Innovates 2026 had roughly 3,000 submissions, each usually a PDF or PPTX link, and a shortlist deadline. the system needed to classify
- [beautiful enough to build](https://a3ro.dev/blog/beautiful-enough-to-build): some ideas make you keep returning to them after the excitement has worn off. not because they are guaranteed to work, but because the world
- [context engineering, or: how i got consistent art out of a diffusion model without training anything](https://a3ro.dev/blog/context-engineering-consistent-art-without-training): the first few generated panels were good enough to make me optimistic and inconsistent enough to make that optimism useless. the character c
- [the machine that thinks but doesn't feel](https://a3ro.dev/blog/the-machine-that-thinks-but-doesn't-feel): i spend a lot of time building with systems that can write, reason through a bug, and occasionally surprise me. the surprising part is not t
- [how to know if you're delusional (no, really)](https://a3ro.dev/blog/reality-vs-delusion): certainty has a similar feeling whether it is earned or not. that is the uncomfortable bit. feeling sure is not evidence that you are right.
- [on b.tech, fake dreams, and the life nobody asked for](https://a3ro.dev/blog/btech-fake-dreams-life): engineering is a good path for some people. the problem starts when it becomes the answer before anyone has had time to ask the question. in
- [on adulting in india — loneliness, guilt, choices nobody trained us for](https://a3ro.dev/blog/adulting-in-india): the strange part of growing up in india is that you can be given a very detailed script for twenty years and then be expected to improvise t
- [agentic ai — the wave i didn’t ask for (but can’t ignore)](https://a3ro.dev/blog/agentic-ai-wave): the word “agentic” is now attached to almost every AI product. sometimes it means a model can use tools and retry a task. sometimes it mostl
- [love, after it stops feeling safe](https://a3ro.dev/blog/love-a-feeling-i-wish-to-never-experience-ever-again): i wrote before about finding someone who feels like home. i meant it when i wrote it. i also know now that a sentence can be true for one ve
- [on having a good partner](https://a3ro.dev/blog/finding-someone-who-feels-like-home): after a day that has taken more than it gave, a good relationship can be surprisingly ordinary: a call, a stupid meme, a message asking whet
- [burnout, comparison, and doing work you can return to](https://a3ro.dev/blog/burnout-bullshit-being-human): some days the code editor feels heavier than it should. not because the problem is impossible, but because you opened the internet first and
- [so you're thinking of learning to code?](https://a3ro.dev/blog/learning-to-code-vibe-era): the first useful thing i ever made with code was small enough to be embarrassing. it saved me from doing the same boring task by hand again.

## Canonical Links & Contact
- Website: https://a3ro.dev
- Location: Prayagraj, Uttar Pradesh 211001, India
- LinkedIn: https://www.linkedin.com/in/akshat-singh-kushwaha/
- GitHub: https://github.com/a3ro-dev
- Twitter/X: https://twitter.com/a3rodev
- Email: akshatsingh14372@outlook.com
- Schedule: https://calendly.com/a3ro-dev/let-s-talk
- bits&bytes: https://gobitsnbytes.org

## Optional
- [Full knowledge base with every essay inlined](https://a3ro.dev/llms-full.txt)
- [RSS feed](https://a3ro.dev/feed.xml)
- [Sitemap](https://a3ro.dev/sitemap.xml)