Manovah Benny
Get in touch
Available for work · Bangalore, IN

Engineering systems
that learn.

I'm Manovah — an independent AI engineer and backend developer shipping real-time ML pipelines and agentic systems for startups and product teams.

View selected workTalk to me
Focus
Production AI
Type
Freelance · remote
Stack
Rust · Python · TS
alpha-engine · tokio · ws://solana
live · sub-ms feature extraction14,238 snapshots
Selected work · 03

Three systems, shipped to production.

01 · Real-time ML

Alpha Engine — a closed‑loop ML pipeline on Solana.

A self‑improving infrastructure that captures 100+ high‑dimensional feature snapshots per signal and auto‑labels outcomes against the on‑chain ground truth.

  • Rust + Tokio ingestion processing thousands of WebSocket events per second, sub‑millisecond feature extraction.
  • Background labeling engine tracking each token at +30m / +1h / +2h, auto‑generating WIN/LOSS labels from peak‑mcap and drawdown.
  • Transitioning rule‑based heuristics to a Gradient Boosted Decision Tree targeting a 20%+ precision lift.
RustTokioSolanaSQLite · WALScikit‑LearnGBDT
Signal stream · auto‑labeled+30m / +1h / +2h
02 · Agentic RAG

An autonomous research agent that cites its sources.

An AI assistant built on Mastra AI and Next.js that searches the web and academic sources, conducts a literature survey, and returns a referenced report.

  • Vector‑store memory for contextual retrieval, backed by file embeddings and Turso (SQLite) for production persistence.
  • Session and user state managed through MongoDB with serverless API routes.
  • Multi‑step planner that interleaves search, read, and cite — and surfaces every claim with a reference.
Mastra AIGoogle ADKNext.jsTursoMongoDBRAG
03 · NLP at the edge

A toxicity detector that moderates communities live.

A deep‑learning classifier trained in TensorFlow, wired into a Telegram bot backend that flags and removes harmful messages in real time.

  • Custom NLP model trained on community‑specific data, exported and served from Node.js for low‑latency inference.
  • Tunable thresholds per community, with full audit trail of every flagged message.
  • Shipped as a drop‑in moderation layer for crypto and developer communities running on Telegram.
TensorFlowNode.jsTelegraf.jsNLP
Agentic UI

Don't want to scroll? Just ask.

Hi — I'm an AI agent answering on Manovah's behalf. Ask me about his projects, technical stack, skills, availability, or how to contact him. I can answer directly or render live interactive cards!

Powered by CopilotKit

Experience · 03

Building backends & bots in production.

2024

Backend Developer

Ozone Chain · Remote

Built a Node.js + TypeScript Telegram bot on Telegraf.js handling user referrals, on‑chain reward withdrawals, and a community tip system. Persisted user data and transaction history through MongoDB / Mongoose.

2023 – 2024

Fullstack Developer

Codinary · Remote

Developed and shipped React + Next.js applications with Node.js backends, integrating Firebase and Supabase. Tuned reliability and performance across production deployments.

2023 — now

Freelance Developer

Telegram bots & web systems

Custom Telegram automation and crypto‑integrated applications for multiple clients — Node.js, Telegraf.js, Next.js. Shipped end‑to‑end, from spec to deploy.

Stack

The toolkit I reach for.

AI & ML
TensorFlowJAXScikit‑LearnMastra AIGoogle ADKRAGVector storesNLPAgentic AI
Languages
PythonJavaScriptTypeScriptRust
Backend & frameworks
Node.jsExpressTokioTelegraf.jsNext.jsWebSocketsBullMQRESTMicroservices
Data & infra
PostgreSQLMongoDBMySQLSQLite (WAL)TursoRedisDockerPM2CI/CDJWT
About

Curious about machines that get better the longer they run.

I'm an independent AI engineer working out of Bangalore. Most of what I do sits at the intersection of machine learning and backend systems: real‑time pipelines, agentic applications, and tooling that keeps a model honest after it ships.

I'm drawn to closed‑loop systems — the kind where the world quietly labels its own data, and the next training run is a little better than the last. I write in Rust when latency matters, Python when ideas matter, and TypeScript when both do.

Outside of client work I read about interpretability, write small tools for myself, and tinker with crypto infrastructure.

Currently

Shipping a real‑time ML alpha engine.

Rust + Tokio ingest, GBDT pipeline, sub‑millisecond features on Solana.
Engagements
  • Production AI featurescontract
  • Real‑time backendscontract
  • Agentic systems / RAGcontract
  • Response within 24hIST, weekdays
Taking on freelance engagements · 2026

Have a system that should think for itself? Let's build it.

I'm currently taking on a small number of focused AI and backend engagements at a time — production ML, agentic systems, real‑time pipelines.

manovah04@gmail.comLinkedIn
Emailmanovah04@gmail.com
Phone+91 87927 09434
BasedBangalore, India · IST (UTC+5:30)
HoursReplies within 24h, weekdays