Sameer Bhalerao
All projects

Community trust infrastructure

CircleWorks

Trust infrastructure for the informal local economy inside residential societies — workers build a reputation that travels, residents find reliable help.

Built + deployed to Railway · pilot-ready at Blueberry Homes, Bangalore

The problem

The informal labour market inside residential societies — plumbers, electricians, cleaners — runs entirely on word-of-mouth: no trust signals, no transparent pricing, no accountability. Residents cannot find reliable workers; good workers cannot build a reputation that travels.

How it works

  • A resident posts a job by typing or speaking — in English, Hindi, Kannada, or Marathi, using browser-native speech recognition. A natural-language parser extracts the structured details — skill, duration, headcount.
  • An LLM pricing engine (Claude) suggests a fair rate range from category, location, and historical rates.
  • Matched workers receive Telegram alerts, accept with one tap, complete the work, and build a verifiable trust score from completions and reviews.
  • All payments are peer-to-peer via UPI — the platform never touches money.
  • Built on Django 5.1 + PostgreSQL on Railway, with a Telegram bot for worker onboarding and notifications, plus full SOPs, job-lifecycle, and dispute-resolution design authored with an on-ground partner.

Architecture

Resident posts a job

type or speak · 4 languages

AI pricing & matching (Claude)

NLP job parserAI rate suggestionSkill / location match
alert via Telegram bot

Worker

One-tap accept

Complete work

Verifiable trust score

Payment direct, peer-to-peer (UPI)

the platform never touches money

Backend

Django 5.1PostgreSQLTelegram Bot APIRailway

Resident posts (type or speak, 4 languages) → parsed & priced → workers matched via Telegram → trust score; money flows P2P, never through the platform.

The journey

Designed for how informal work actually happens in residential societies — no formal contracts, no app-to-app coordination. Word-of-mouth trust, now with a structure underneath it.

01

The resident types or speaks what they need — in their own language

They open the app and describe the job exactly as they'd tell a neighbour — "Need someone to water the plants this Saturday" — by typing or by voice, in English, Hindi, Kannada, or Marathi. The system reads it, extracts the structure, and estimates a fair price, all before they confirm anything.

Plain English in — AI reads, extracts, and prices before you confirm

02

The LLM pulls job structure from the description automatically

Skill, date, duration, headcount — extracted without a dropdown or a form field. The resident reviews what the AI understood, corrects anything if needed, and confirms. They never touched a category picker.

LLM extracts skill, date, duration, headcount — no form filled

03

Claude sets a specific, justified price

Not a range to negotiate down from. A concrete number — ₹200 — derived from job category, duration, and local market rates for that society. Residents don't haggle. Workers know what to expect before they apply.

A specific price, not a range — derived from category and market rates

04

The right workers in the society get a Telegram alert — no new app to install

Workers matched by skill and availability receive a job notification in the app they already have open. The platform never pushed them to download anything new. One tap to see the details, another to apply.

Job alert lands in Telegram — apply without leaving the chat

Full schedule management via bot commands — /online, /offline, /jobs

05

ID-verified workers are browseable — skills and availability visible before you hire

The Pros directory shows only ID-claimed professionals from your society. The resident can see who they're dealing with — skill tags, availability windows, tenure — before committing to anyone.

Verified pros from your society — ID on file, skills tagged, available to hire

Highlights

  • Voice + text in 4 languages
  • Natural-language job parsing
  • LLM pricing engine
  • Telegram-native
  • Peer-to-peer payments

Stack

Django 5.1PostgreSQLTelegram Bot APIClaude APIRailwayPython

Why it matters

A marketplace designed for trust, not transactions — with the operational design (SOPs, disputes, field ops) that a real pilot needs, not just the code.