━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
SUPPLEMENTARY FIELD NOTES · FILE NO. 2005-ANB-019 · PAGE 2
DEA SOUTHWEST DIVISION — BENGALURU ATTACHÉ OFFICE
ANALYST COPY · DO NOT DUPLICATE · DESTROY AFTER REVIEW
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
┌─────────────────────────────────────────────────────────────────┐
│ SUBJECT : Anubhab Pradhan · @ANUBprad │
│ LOCATION : Bengaluru, Karnataka, IN │
│ CURRENT COVER : B.E. student · CMRIT · Final year │
│ REAL OCCUPATION : Building production LLM systems │
│ │
│ 🔴 ACTIVELY BUILDING — New operation in progress │
│ 🟡 PLACEMENT WINDOW — Opens August 2026 │
│ 🟢 OPEN SOURCE — PRs being filed as you read this │
│ │
│ AVAILABILITY : Full-time from June 2027 │
│ OPEN TO : AI-native startups · FAANG · Research │
│ WILL NOT DO : Tutorial projects · Demo-only roles │
└─────────────────────────────────────────────────────────────────┘
| OPERATION | CLASSIFIED BRIEF |
|---|---|
| RedOps | Autonomous LLM red-teaming network. Six attack vectors. 100+ live payloads. Scores itself. Nobody pulls the trigger twice. |
| Kairos | Four-agent orchestration cell. Planner → Retriever → Reasoner → Validator. Self-operating. Running unsupervised since deployment. |
| APEXiq | Natural language to SQL pipeline. Agents write queries, validate them, report back. F1 telemetry. Business intelligence. No human in the loop. |
| DeepDive | Video intelligence system. Transcribes, chunks, indexes, retrieves. Bilingual. Runs on free APIs. Hours of footage in seconds. |
Lamatic.ai — Applied AI Engineer (Jan – May 2026)
Built LLM agent pipelines for end-to-end workflow automation. Workflows that required operators stopped requiring them. Subject made himself structurally irreplaceable, then moved on.
Zidio Development — Data Science Intern (Aug – Nov 2025)
Stacked ARIMA, Prophet, LSTM into single ensemble. Built automated evaluation pipeline. Team inherited a system that worked harder than the people before it.
The following instruments were recovered from subject's workstation. All are operational. All are dangerous in the right hands.
Languages & Frameworks
Backend & Web
Databases & Caching
DevOps & Infrastructure
Cloud & Systems
Data Science & ML
LangChain Infiltrated core repository · Issue #31802
Metaflow Embedded in production ML pipeline work
AOSSIE OpenVerifiableLLM · Found critical bugs
Real-time activity tracking. Commits logged. Patterns analyzed. Subject stays active.
TOTAL CONTRIBUTIONS : 229 | CURRENT STREAK : 4 days | LONGEST STREAK : 12 days
────────────────────────────────────────────────────────────────────────────────────
LAST ACTIVITY : TODAY | OPERATIONAL STATUS : ACTIVE | THREAT LEVEL : HIGH
┌──────────────────────────────────────────────────────┐
│ ACTIVITY LEVEL : EXTREMELY HIGH │
│ COMMIT FREQUENCY : DAILY · MULTIPLE │
│ DEPLOYMENT VELOCITY : RAPID · CONSISTENT │
│ LAST KNOWN ACTIVITY : TODAY │
│ OPERATIONAL STATUS : ACTIVE · BUILDING │
└──────────────────────────────────────────────────────┘
INSTITUTION : CMR Institute of Technology, Bengaluru
PROGRAM : B.E. Artificial Intelligence & Data Science
ENROLLED : September 2023
TARGET DATE : June 2027
FILE ID : 1CR23AD019
Filed by Field Agent [REDACTED] · 25 May 2005
"Subject does not wait for problems. He finds gaps, builds tools, ships them. His red-teaming runs in production right now. His agent networks operate without human oversight by design.
We have not been able to determine the full extent of ████████████████████████████████████████████████.
Do not approach without technical backup. Do not underestimate the intern."
— Walter White · Breaking Bad · S05E16