AI & Data Infrastructure Engineer

I build production systems for AI automation, data pipelines, and analytics.
Backend architecture · LLM orchestration · Scalable data infrastructure.

I'm Ayush Verma — a software engineer with ~3 years of hands-on experience building end-to-end AI and data systems. Everything I ship is real, Dockerized, and deployed.

Explore selected systems and case studies below.

LangGraph
LangChain
VertexAI
RAG / FAISS / pgvector
OSS LLMs (Ollama, vLLM)
Whisper (STT)
HMM & Regime Detection
Scikit-learn
Pandas / NumPy
Hugging Face
Python
FastAPI
PostgreSQL
TimescaleDB
Redis
SQLAlchemy
Pydantic
WebSockets
Prometheus
React
Next.js
TypeScript
Tailwind CSS
Framer Motion
Three.js
React Three Fiber
TradingView Charts
Docker
GitHub Actions
GCP / Azure
Cloud Computing
DigitalOcean / Railway
Linux / Shell Scripting
Process Automation
Webflow
Bubble
Airtable
Zapier / Make
Softr / Glide
Statistical Arbitrage
Walk-Forward Backtesting
Binance WebSocket API
Cointegration Analysis
Risk & Monitoring
Event-Driven Architecture
LangGraph
LangChain
VertexAI
RAG / FAISS / pgvector
OSS LLMs (Ollama, vLLM)
Whisper (STT)
HMM & Regime Detection
Scikit-learn
Pandas / NumPy
Hugging Face
Python
FastAPI
PostgreSQL
TimescaleDB
Redis
SQLAlchemy
Pydantic
WebSockets
Prometheus
React
Next.js
TypeScript
Tailwind CSS
Framer Motion
Three.js
React Three Fiber
TradingView Charts
Docker
GitHub Actions
GCP / Azure
Cloud Computing
DigitalOcean / Railway
Linux / Shell Scripting
Process Automation
Webflow
Bubble
Airtable
Zapier / Make
Softr / Glide
Statistical Arbitrage
Walk-Forward Backtesting
Binance WebSocket API
Cointegration Analysis
Risk & Monitoring
Event-Driven Architecture
LangGraph
LangChain
VertexAI
RAG / FAISS / pgvector
OSS LLMs (Ollama, vLLM)
Whisper (STT)
HMM & Regime Detection
Scikit-learn
Pandas / NumPy
Hugging Face
Python
FastAPI
PostgreSQL
TimescaleDB
Redis
SQLAlchemy
Pydantic
WebSockets
Prometheus
React
Next.js
TypeScript
Tailwind CSS
Framer Motion
Three.js
React Three Fiber
TradingView Charts
Docker
GitHub Actions
GCP / Azure
Cloud Computing
DigitalOcean / Railway
Linux / Shell Scripting
Process Automation
Webflow
Bubble
Airtable
Zapier / Make
Softr / Glide
Statistical Arbitrage
Walk-Forward Backtesting
Binance WebSocket API
Cointegration Analysis
Risk & Monitoring
Event-Driven Architecture
LangGraph
LangChain
VertexAI
RAG / FAISS / pgvector
OSS LLMs (Ollama, vLLM)
Whisper (STT)
HMM & Regime Detection
Scikit-learn
Pandas / NumPy
Hugging Face
Python
FastAPI
PostgreSQL
TimescaleDB
Redis
SQLAlchemy
Pydantic
WebSockets
Prometheus
React
Next.js
TypeScript
Tailwind CSS
Framer Motion
Three.js
React Three Fiber
TradingView Charts
Docker
GitHub Actions
GCP / Azure
Cloud Computing
DigitalOcean / Railway
Linux / Shell Scripting
Process Automation
Webflow
Bubble
Airtable
Zapier / Make
Softr / Glide
Statistical Arbitrage
Walk-Forward Backtesting
Binance WebSocket API
Cointegration Analysis
Risk & Monitoring
Event-Driven Architecture

Selected Projects

Real systems, shipped and deployed.

CRM / Retention / Winback Strategy Engines
View Details

CRM / Retention / Winback Strategy Engines

Benchmark
Live

Lifecycle automation engines for conversion, retention, and revenue recovery

full stack
Python / FastAPI
PostgreSQL
+4
View Details

Mangal Murti Jeweller — Diamond E-Commerce Platform

Benchmark
Delivered

Full-stack luxury diamond marketplace with 176K+ certified stones

full stack
Next.js
React / TypeScript
+5
View Details

DocIntel v2 — Document Intelligence Platform

Benchmark
Delivered

End-to-end RAG pipeline with live ingestion, retrieval, and evaluation

ai
Python / FastAPI
React / TypeScript
+4
Divergence Detector (Advanced)
View Details

Divergence Detector (Advanced)

Benchmark
Live

Event-driven market scanner processing 500+ ticks/s

trading
Python / FastAPI
React / Tailwind
+4

Code Snippets

Small excerpts from real systems — the kind of details that make products reliable.

Stable content hashing + near-duplicate detection

A layered dedupe strategy: stable hash → pg_trgm similarity → pure-Python fallback.

Python
1# normalize + stable hash (accounts for platform & angle)
2def normalize_content(content: str) -> str:
3 normalized = content.lower()
4 normalized = re.sub(r"\s+", " ", normalized)
5 normalized = re.sub(r"[^\w\s]", "", normalized)
6 return normalized.strip()
7
8
9def compute_content_hash(content: str, platform: str, angle: str) -> str:
10 normalized = normalize_content(content)
11 hash_input = f"{normalized}|{platform.lower()}|{angle.lower()}"
12 return hashlib.sha256(hash_input.encode("utf-8")).hexdigest()
13
14
15# orchestrator: exact-hash → trigram → fallback
16async def check_duplicate(session, campaign_id, content, platform, angle, threshold=0.7):
17 content_hash = compute_content_hash(content, platform, angle)
18
19 exact = await check_exact_duplicate(session, campaign_id, content_hash)
20 if exact:
21 return {"is_duplicate": True, "type": "exact", "existing_post_id": exact.id}
22
23 near = await check_near_duplicate_trgm(session, campaign_id, content, threshold)
24 if not near:
25 near = await check_near_duplicate_fallback(session, campaign_id, content, threshold)
26
27 return {"is_duplicate": bool(near), "type": "near" if near else None, "content_hash": content_hash}

JSON-LD graph injector

Keeps schema composable while emitting one canonical @context (and @graph when needed).

TypeScript / React
1type JsonLdProps = {
2 data: Record<string, unknown> | Array<Record<string, unknown>>;
3};
4
5export function JsonLd({ data }: JsonLdProps) {
6 const payload = Array.isArray(data) ? data : [data];
7 const graphNodes = payload.map((node) => {
8 const { "@context": _ctx, ...rest } = node as Record<string, unknown>;
9 return rest;
10 });
11
12 const json =
13 graphNodes.length === 1
14 ? { "@context": "https://schema.org", ...graphNodes[0] }
15 : { "@context": "https://schema.org", "@graph": graphNodes };
16
17 return (
18 <script
19 type="application/ld+json"
20 dangerouslySetInnerHTML={{ __html: JSON.stringify(json) }}
21 />
22 );
23}
Skills & Capabilities

Stack I actually use.

These are the tools in the projects above — Python, FastAPI, LangGraph, React/Next.js, PostgreSQL, Redis, Docker, and the surrounding ecosystem.

AI & LLM

10 skills

LLM orchestration, RAG pipelines, and applied ML.

LangGraphLangChainVertexAIRAG / FAISS / pgvectorOSS LLMs (Ollama, vLLM)Whisper (STT)HMM & Regime DetectionScikit-learnPandas / NumPyHugging Face

Backend & Data

9 skills

APIs, async pipelines, and data persistence.

PythonFastAPIPostgreSQLTimescaleDBRedisSQLAlchemyPydanticWebSocketsPrometheus

Frontend

8 skills

Production UIs and interactive dashboards.

ReactNext.jsTypeScriptTailwind CSSFramer MotionThree.jsReact Three FiberTradingView Charts

Infra & Deployment

7 skills

End-to-end deployment and cloud infrastructure.

DockerGitHub ActionsGCP / AzureCloud ComputingDigitalOcean / RailwayLinux / Shell ScriptingProcess Automation

Low-Code & Automation

5 skills

Rapid delivery with robust integrations.

WebflowBubbleAirtableZapier / MakeSoftr / Glide

Quant & Data Systems

6 skills

Market data pipelines and strategy tooling.

Statistical ArbitrageWalk-Forward BacktestingBinance WebSocket APICointegration AnalysisRisk & MonitoringEvent-Driven Architecture

How I Work

I handle architecture, backend, and delivery myself — end to end, Dockerized, and deployed. For larger scopes, I bring in people I've worked with before. You deal with one person throughout.

Trusted to Ship Fast

Real feedback from clients on Upwork

Exceptional developer. Delivered a 30-day scope in 5 days with clean architecture and major infrastructure cost optimization. Proactive, strategic, and explains systems clearly. Highly recommended.

SV

Sahil V.

ClientUpwork

Working with him was honestly a great experience. He went above and beyond what I expected, the whole system runs super smoothly, and you can tell he's put a lot of thought into how everything's built, especially the backend and infrastructure. What stood out to me was how smartly he optimized everything. Instead of adding extra paid tools, he used open-source options and self hosted n8n, which saved costs without cutting corners. I really appreciated that level of practicality. Took the time to hop on a call and walk me through the entire setup, shared clear documentation, and made sure I understood. Overall, he's easy to work with, communicates clearly, and genuinely cares about doing quality work. I'd definitely love to work with him again.

SV

Sahil V.

ClientUpwork

Frequently Asked Questions

Everything I ship is real, Dockerized, and deployed end-to-end. Common questions about how I work and what to expect.

Let's Build Something Great

I'm always interested in working on exciting projects and collaborations.

Chat on WhatsApp