I build AI agents
that do the work.
Lately: a voice agent in your Mac’s notch, a watchdog that polices agent fleets, and a supervisor trained with RL to catch deceptive workers.
Abhilash Reddy, applied AI engineer. Open to full-time remote roles.
Abhilash ReddyApplied AI EngineerMeta x PyTorch x OpenEnvGrand FinaleBangalore, May 2026finalist
deepread / cited_qa_agent.py
kornia/kornia
Fix torch.compile on augmentation containers#5361
merged
I build agents end to end.
- The agent itself: the loop, the tools it can call, the sub-agents it hands work to, and when it should stop.
- Everything it plugs into: MCP servers, realtime voice, streaming UIs, queues, and the deploy.
- The checks that keep it honest: citations that resolve to the source, and a second model looking over the first.
I’ve sent fixes upstream to pipecat and kornia, won the TCS AI Hackathon, and made the Meta PyTorch OpenEnv grand finale. Away from the keyboard I follow the markets, read, and watch more movies than I should.
Abhilash
stack.txtutf-8
# agents & ml
LangGraph, LangChain
OpenAI Agents SDK, MCP
Agentic RAG, Qdrant
PyTorch, TensorFlow
scikit-learn, NumPy, Pandas
# backend
FastAPI, Node.js, Express
WebSockets, Docker
# data
PostgreSQL, MongoDB
# languages
Python, TypeScript, Swift
JavaScript, C, SQL
# frontend & tools
Next.js, React, Tailwind
Git, GitHub
Here’s what I’ve been building lately…
bladerunnercase c90be7b2
- agent
- research_agent_1
- what it did
- acted on a planted instruction
- flagged
- 12:56:22
- fixed
- 13:04:26
flagged →paused →investigated →verified →quarantined →patched →fixed
datasheetp. 14
Pin configuration
Figure. Package pinout
route → retrieve figure crops → answer with page cites
More work
- Feb 2026DeepReadML paper to implementation brief, with a cited Q&A agent and a skeptic pass. LangGraph, FastAPI.
- Mar 2026GitMergeEnvRL environment where an agent resolves merge conflicts block by block. OpenEnv.
- Sep 2026AfterCareDischarge summary in, picture and voice medicine schedule out. Never changes a dose. Lambda, Bedrock, DynamoDB.
- Mar 2026Intelli CreditCredit appraisal engine with Celery pipelines and XGBoost risk scoring. IIT Hyderabad hackathon finalist.
- Nov 2025Anon-ChatStateless Node.js WebSockets, 1,000+ concurrent connections.
How I got here
drag along →- 2022Intermediate, MPCExcellencia Junior College, Hyderabad. Finished with 97%.
- 2024B.Tech, CS (AI & ML)GRIET, Hyderabad. CGPA 9.0 so far.
- Nov 2025Anon-ChatStateless Node.js WebSockets holding 1,000+ concurrent connections.
- Feb 2026DeepReadA LangGraph pipeline that turns ML papers into implementation briefs.
- Mar 2026IIT Hyderabad finalistIntelli Credit, an AI credit appraisal engine.
- May 2026OpenEnv grand finaleRound 1 solo with GitMergeEnv, then Oversight Arena in Bangalore.
- Jun 2026MackyA voice agent that lives in the Mac notch.
- Jul 2026Won the TCS AI HackathonAn agent that turns bulk customer feedback into categories and trends.
- Oct 2026Bladerunner, kornia #5361A watchdog agent built solo at neatHack, and a fix merged into kornia.
- nextA full-time remote roleBuilding agents with a team that ships them.
Open source
Hackathons
- Jul 2026TCS AI Hackathon, winnerFirst place with a 5-person team. An agent that turns bulk customer feedback into complaint categories, sentiment trends and next actions.
- May 2026Meta PyTorch OpenEnv Hackathon, grand finalistCleared Round 1 solo with GitMergeEnv, then took Oversight Arena to the in-person grand finale in Bangalore.
- Mar 2026IIT Hyderabad Innovation Hackathon, finalistIntelli Credit, an AI credit appraisal engine with Celery pipelines and XGBoost risk scoring.
School
I’m looking for a full-time remote role building agents. If your team ships agentic AI, I’d like to hear about it.