2+ years taking GenAI products from whiteboard to deployment at Entelegen Analytics LLP: architecture, backend APIs, containerised deployment on SAP AI Core, GCP and AWS, and client handover.
A GenAI engineer with 2+ years of production experience building LLM, RAG and multi-agent systems for enterprise clients. He takes products from whiteboard to deployment, and measures whether they actually work.
2+ years taking GenAI products from whiteboard to deployment at Entelegen Analytics LLP: architecture, backend APIs, containerised deployment on SAP AI Core, GCP and AWS, and client handover.
RAG pipelines, LangGraph multi-agent orchestration, and turning vague business problems into production AI systems people actually use.
Evaluation. Shipping an LLM system is the easy half; knowing whether the last prompt or model change made it worse is the hard half. That is what I build now.
Every project below went past the demo stage. Open a span to see what I built and with what.
Integrated NLP/ML models into web services; built DataCLI, a Python (Click) CLI that automates preprocessing.
Annotated training data and ran object-detection model experiments.
Co-authored 'Coronary Artery Disease Prediction using ML Algorithms' (IEEE ICESIC 2022).
Catching made-up answers from a RAG system without paying an LLM judge to check every answer: free rules first, a small trained classifier next, and the judge only for what the first two can't settle.
6 rules for prompt injection and jailbreaks, checked on the question
Every answer. Flags hard evidence: a guess after a refusal, a number the excerpts never mention
Logistic regression on 12 cheap signals; only clears answers, never blocks
Checks every claim against the retrieved excerpts · sees 35% of answers
| System | Sent to judge | Caught | Cost / 1k | Typical time |
|---|---|---|---|---|
| Judge every answer | 100% | 100%* | $7.08 | 2.65 s |
| Cascade | 35% | 92% (11/12) | $2.60 | 0.19 s |
| Cascade, length rule instead of classifier | 45% | 92% | $3.27 | 0.10 s |
| Rules only | 0% | 50% | $0 | 0.06 s |
40 held-out answers, 12 of them hallucinations, with thresholds chosen on training data first. 95% intervals: 20–50% sent to the judge, 71–100% caught. *Partly by construction: the test reference uses the judge's label where the human labeller agreed with it.
Test ROC-AUC 0.71 vs 0.80 for answer length, intervals overlapping. Its value is cost: about 10 points fewer judge calls at the same recall.
They reuse its wording while misreading it, so word overlap was useless (AUC 0.52). Embeddings measure topic, not truth.
60 answers labelled blind: 67% agreement (κ 0.25) at first, mostly from inconsistent refusal labels. Reported as is, not hidden.
19 of 24 test attacks blocked; all 5 reworded attacks written afterwards got through.
A personal automation server in Docker with persistent workflows: the same low-code tool I shipped to a client as an alternative to a hand-built agent.
Hover a cluster to see its neighbours. Placed by hand, not by an embedding model.
Aug 2026 – Aug 2028
SRM Institute of Science and Technology · 2019–2023 · CGPA 8.9/10
IEEE ICESIC 2022
# open to GenAI / LLM / AI Engineer roles await adhith.contact({ email: "adhithsagit0510@gmail.com", github: "github.com/CodeSavitar", based_in: "Coimbatore, India", })