Skip to content
Manohar Papasani Get in touch

AI R&D lab · Hyderabad

A digital workforce that runs itself

I'm Manohar Papasani, AI Developer Lead at Minfy. In my R&D lab, 36 AI agents research, write, check and publish four posts a day on five platforms, and I measure every step.

See how it works Get in touch
Digital workforce Live ClockFour slots a day Agent teamPlanner, scouts, writer CriticFour judges LinkedInThen four more platforms Each team has its own clock, so onefailure never stops another. Agents research and write in myvoice, from verified facts only. A weak post is never published; theslot re-runs and tries again. X, Bluesky, Facebook and Dev.tofollow at the same times every day.
36 AI agents in production
20 Posts a day across five platforms
28 Awards at Bank of America, including Platinum
14,300 Engineer-hours saved a year at Bank of America

The workforce

Nine teams, each on its own clock

Meet all 36 agents

What the lab builds

Systems that do real work, and the research behind them

Live every day

Digital workforce

Nine agent teams research, write, check and publish on their own. Each has its own schedule, repairs its own missed runs and is tested every week.

See how it works Open research

Can an open model become a reliable coding agent?

A study of Gemma 4 on real software tasks, with planted bugs that test whether the tester can be fooled.

Read the research Free for everyone

13 free courses

From Python to AI security, MCP and LLM evaluation, built for engineers who learn by doing.

Browse the courses

Measured, not claimed

The workforce grows the brand it writes for

The agents started posting on 11 Oct 2026. My analytics agents collect these numbers from LinkedIn twice a week, and code rejects any figure that is not in the data.

Followers
992
Impressions, past year
69,506
Posts by the agents
54

Updated 3 Nov 2026. Sources: LinkedIn analytics, my public profile and our posting records.

See the full analytics

Impressions per month Last 12 months
5k 10k Dec 2025: 0 impressions Dec Jan 2026: 75 impressions Jan Feb 2026: 10,595 impressions Feb Mar 2026: 7,190 impressions Mar Apr 2026: 1,417 impressions Apr May 2026: 6,231 impressions May Jun 2026: 2,096 impressions Jun Jul 2026: 14,883 impressions Jul Aug 2026: 12,912 impressions Aug Sep 2026: 584 impressions Sep Oct 2026: 11,298 impressions Oct Nov 2026: 2,221 impressions Nov Agents started

Lab notebook

Latest from the lab

Read all 27 entries
  1. Shipped

    Each team now has its own scheduler, and each agent its own test

    The five agent teams used to share one scheduler, so a single mistake could delay every team.

  2. Shipped

    Every post now reaches five platforms, on time

    X joined Bluesky, Dev.to and Facebook, and every post now ends with the same five follow links.

  3. Shipped

    Tracing caught the cost report counting some calls twice

    While switching on tracing for the insights team, its trace showed each Claude Haiku call twice: once from the agent framework and once from a newer version of the model provider's own library.

Career

More than ten years of building software that does real work

I started with automation, data platforms and production systems. Today I build AI agents.

Read the full story

  1. Oct 2026 – present

    Minfy · AI Developer Lead

    I lead a team of four engineers building agentic GenAI solutions on AWS.

  2. Aug – Oct 2026

    ValueMomentum · GenAI Solution Architect

    I designed the agent factory behind a client's 20-agent programme, delivered in six to eight months instead of two years.

  3. Apr 2021 – Jul 2026

    Bank of America · AVP, GenAI and Automation Lead

    I saved about 14,300 engineer-hours a year, founded the prompt-engineering team and won 28 awards in five years.

Follow the lab

The same work, on six channels

My digital employees publish every post on LinkedIn first, then share it on the other channels. Pick the one that suits you.

Back the lab

Build the next digital employees with me

I'm looking for partners, sponsors and early users for the next builds. Each one ships in public, with measured results.

Talk to me on LinkedIn See the code on GitHub

Coming next

  • Agent swarm A factory that turns a short spec into a working agent, and a shared board where 100 agents coordinate.
  • Guardrails agent It interviews a team, then enforces their rules across every agent.
  • Competition autopilot It finds AI competitions worldwide, builds entries and fixes its own failures.
  • Machine learning inside agents Routers, rankers and guards that make agents cheaper and safer.