Akshay Bajpai ax5hay

Akshay Bajpai

In thrust we trust.

I build reliable, robust software systems, AI or otherwise.
The kind that holds under load, states its own assumptions, and fails loud instead of silent.

Thrust is cheap. Controlled thrust is the whole job.

%%{init: {'theme':'base','themeVariables':{'primaryColor':'#15151c','primaryTextColor':'#f3ede0','primaryBorderColor':'#d8bd85','lineColor':'#6f6a5f','fontSize':'15px'}}}%%
flowchart LR
  F(["ideas and raw data"]) --> T["THRUST<br/>ship the thing"]
  T --> C{"control<br/>surfaces"}
  C --> O["orchestration"]
  C --> S["safety and guardrails"]
  C --> D["schema that<br/>refuses bad data"]
  C --> M["observability"]
  O --> R(["TRUST<br/>holds under load, in prod"])
  S --> R
  D --> R
  M --> R
  classDef thrust fill:#8c2f24,stroke:#d8bd85,stroke-width:1.5px,color:#f6f1e6;
  classDef hub fill:#14141a,stroke:#d8bd85,stroke-width:1.5px,color:#d8bd85;
  classDef node fill:#15151c,stroke:#3a3a42,color:#e8e3d6;
  classDef seed fill:#0f0f14,stroke:#3a3a42,color:#b4b0a4;
  class T,R thrust;
  class C hub;
  class O,S,D,M node;
  class F seed;

I design systems the way you'd design a propulsion stack: thrust is easy, control is the work. Most of what I build sits in the unglamorous middle: the orchestration, the safety layer, the schema that refuses bad data, the job queue that survives a restart. Shipping a demo is thrust. Keeping it upright in production is trust.

I work across the stack but gravitate toward AI systems that have to behave: LLM assistance that stays inside guardrails, pipelines that turn mess into structured truth, and vision/edge workloads that run where the data actually lives.

What I'm flying right now

_{A map of what I build, and the domains it falls into.}

%%{init: {'theme':'base','themeVariables':{'primaryColor':'#15151c','primaryTextColor':'#f3ede0','primaryBorderColor':'#d8bd85','lineColor':'#6f6a5f'}}}%%
flowchart TB
  ME(("Akshay<br/>Bajpai"))
  ME --> AI["AI systems<br/>and LLMs"]
  ME --> CV["computer vision<br/>and edge"]
  ME --> DP["data platforms<br/>and pipelines"]
  ME --> WP["web, product<br/>and sites"]
  AI --> AURIXA["AURIXA"]
  AI --> GHDA["GHDA-SaaS"]
  AI --> OCR["OCR-LLM-DIST"]
  CV --> VIG["Vigilix"]
  CV --> DADM["DADM"]
  CV --> FS["FlowState"]
  DP --> AIDA["AIDA"]
  DP --> RS["RestroScraper"]
  WP --> MST["MirrorState"]
  WP --> SITE["akshaybajpai.com"]
  classDef core fill:#8c2f24,stroke:#d8bd85,stroke-width:2px,color:#f6f1e6;
  classDef dom fill:#14141a,stroke:#d8bd85,stroke-width:1.3px,color:#d8bd85;
  classDef proj fill:#15151c,stroke:#3a3a42,color:#e8e3d6;
  class ME core;
  class AI,CV,DP,WP dom;
  class AURIXA,GHDA,OCR,VIG,DADM,FS,AIDA,RS,MST,SITE proj;

AURIXA · care-ops platform _{Multi-tenant conversational care operations: governed LLM assistance over one orchestration and safety layer. TypeScript · Python · Next.js · FastAPI · Fastify}	AIDA · health intelligence _{Postgres → Prisma → NestJS → Next.js analytics with optional LLM narration. The UI never opens a DB connection. TypeScript · Prisma}
DADM · defensive AI mesh _{Edge anomaly detection, federated learning, defense-ontology graph. Offline-first, zero-trust, air-gap capable. Rust · Python · ONNX · Neo4j}	Vigilix · AI video wall _{Live multi-camera wall for security ops: RTSP ingest, YOLO overlays, incident alerts. React · FastAPI · Python}
GHDA · gov health automation _{Turns messy, multilingual survey reports into structured JSON, gap analysis, and compliance insight. FastAPI · Fastify · OCR}	akshaybajpai.com · this person, as a drawing set _{A personal site issued as numbered architectural sheets, re-issuable in three CSS-only states. Next.js 15 · zero 3D runtime deps}

_{More in the repository index: lead CRMs, job-application automation, OCR and local-LLM document chat, a behavioral-evidence OS, and a pile of ML notebooks from earlier flight tests.}

The stack I actually touch

_{Every tool below appears in a shipped repo. Pulled by scanning every dependency manifest across all my repositories.}

Languages
Frontend & UI
Dataviz
Backend & APIs
AI, ML & data science
Computer vision & docs
LLMs, RAG & NLP
Data & storage
Infra, CI/CD & cloud
Observability & security
Testing & quality
Desktop & integrations