AI Engineer · Agentic LLM Systems & Multi-Agent Infrastructure
Author of TrueNorth · I design, build, and ship production-grade AI systems end to end.
I build agentic AI systems — multi-agent pipelines, LLM infrastructure, and on-device inference — and take them all the way to production.
- Open source author — shipped TrueNorth to PyPI & NPM: an LLM infrastructure engine with 1,258 passing tests, a 13-stage safety pipeline, and 8-provider routing (~90% cost reduction).
- Led a 10-person team across frontend, backend, and mobile — delivering two concurrent AI product lines.
- Hackathon participant — SANS FIND EVIL! Hackathon (DFIR Automation Track) · Google Cloud Rapid Agent Hackathon (GitLab Partner Track).
- Research-grade rigor — published benchmark results (100% precision on SRL-2018 APT data), fine-tuned models on Hugging Face, experiment tracking on W&B.
- Based in Bengaluru, India · Open to remote-first AI engineering roles (IST, comfortable with US/EU overlap).
| Project | What it does | Stack | Highlights |
|---|---|---|---|
| TrueNorth | Developer-first LLM infrastructure engine — declare the outcome in YAML, it owns the full multi-turn conversation lifecycle | Python · TS · Go · RN | 1,258 tests · 4 SDKs · hallucination firewall (94%) · 8-provider routing · PyPI + NPM |
| ShiftLeft | Autonomous 5-agent bug-fixing pipeline: reads repo → triages → generates fix → opens MR | Python · LangGraph · Gemini · GitLab MCP | End-to-end in ~60s, zero human steps · Google Cloud Rapid Agent Hackathon |
| LogPoseSIFT | Autonomous DFIR orchestrator — MCP server wraps 200+ SANS SIFT tools as typed Go endpoints | Go · Claude · Gemini · MCP · Volatility 3 | 100% precision · 92.8% recall · 0 hallucinations · SANS FIND EVIL! Hackathon |
| PocketLLM | 100% offline Android AI chat running LLMs on-device via MediaPipe C++ bridge | React Native · Expo · MediaPipe C++ · AWS S3 | 9 open-weight models (0.4–5.2 GB) · prompts never leave device |
| AxisMapper | Open-source ICD-10 medical classification & insurance-intelligence model | Fine-tuned LLM · Hugging Face | Published model · medical coding automation |
| AtomicRAG | Multi-hop question decomposition into atomic sub-queries with dependency graphs for RAG | Python · Qwen2.5 · RAG | Fine-tuned retrieval pipeline for complex queries |
Fine-tuned models live on Hugging Face · Training runs tracked on Weights & Biases
| Submission | Hackathon | Description |
|---|---|---|
| ShiftLeft | Google Cloud Rapid Agent Hackathon · GitLab Partner Track | Autonomous GitLab bug-fixing agent — label an issue, autonomous 5-agent pipeline reads the repo, triages the bug, writes the fix, and opens an MR in under 60 seconds |
| Poneglyphs — ShiftLeft | Google Cloud Rapid Agent Hackathon · GitLab Partner Track | Label a GitLab issue shiftleft → 5-agent pipeline reads GitLab Orbit, triages bug, writes fix, opens MR |
| LogPoseSIFT | SANS FIND EVIL! Hackathon · DFIR Automation Track | Autonomous DFIR orchestrator — deploys a specialized AI crew via strict MCP endpoints, executing SIFT diagnostics to triage and self-correct in seconds |
| AllBlue | SANS FIND EVIL! Hackathon · DFIR Automation Track | Splunk alerts trigger autonomous AI forensic triage — findings pushed back as structured IOC events. 100% precision, 0 hallucinations. Claude + Go MCP + SIFT |
Fine-tuned and published models at huggingface.co/AmareshHebbar
| Model | What it does |
|---|---|
| AxisMapper | ICD-10 medical classification and insurance intelligence — fine-tuned for medical coding automation |
| (more models being published) | Actively publishing fine-tuned models for RAG, query decomposition, and domain-specific tasks |
Training runs tracked on Weights & Biases
- Published TrueNorth to PyPI and NPM (Apache 2.0)
- 1000+ problems solved on LeetCode
- B.E. Computer Science & Engineering, Dayananda Sagar College of Engineering (2021–2025)
Open to remote-first AI engineering roles — LLM infrastructure, agentic systems, or AI product engineering.
Let's build something intelligent.