Senior Software Engineer, AI/ML · ValGenesis Chennai, India

Models are probabilistic.
The systems I build around them aren't.

I build production LLM and agentic systems, and the secure multi-tenant backends under them, for regulated domains: legal, life sciences, fintech. Intelligence sits inside the workflow; every output is verified, cited and measured.

The model decides Checks verify A human approves Every change is measured
~70% → ~90%
large-document extraction accuracy, gold-set measured
ValGenesis
weeks → ~1 day
to ship a new agent on a reusable framework
Enegma
~30% faster
Venmo monolith-to-microservices migration, agentic
PayPal
60M+ records
migrated with zero data loss, 15% ahead of schedule
Kissflow
01 · Approach

Not if-else. Intelligence.

Hard-coded workflows break on the case nobody wrote. I put a model in the loop that reads each case, decides what to do, checks its own work, and hands off to a person when it isn't sure. Pick a document and run both.

02 · Large-document workflows

Where the intelligence sits

Long, regulated documents. At every step the model makes a judgment, and every judgment is verified, cited to its page, and measured against a gold set.

03 · Use cases

Problems I've solved with agents

What the agents do, and what changed.

04 · How I work a problem

Measure first. Change the system. Prove it.

AccuracyValGenesis
Problem

Extraction stuck at ~70%

Long regulated documents

Diagnose

Gold-set regression harness

Measure before tuning

Move

Rebuilt evidence retrieval

Plus targeted model tuning

Result
~90%

extraction accuracy

DataKissflow
Problem

60M+ records to move

GCS → MongoDB

Diagnose

Zero loss, faster reads

Two hard constraints

Move

Batch + optimized indexing

Batch processing

Result
Zero loss

15% ahead of schedule · retrieval time −40%

BackendSaama
Problem

Slow REST APIs

Two clinical SaaS products

Diagnose

Profiled the hot paths

Queries and data model

Move

Query optimization

Data modeling

Result
+20%

API efficiency

05 · Where I've shipped

Five roles, one direction

Backend and data first, then retrieval and agents, and now the controls that let AI run in regulated work.

Toolkit

Agents & LLMs

LangGraphLangChainPipecatmulti-agent orchestrationprompt engineeringmicro-promptingcontext cachingthinking-budget tuningGeminiAzure OpenAI

Retrieval

RAGgraph-RAGhybrid searchre-rankingembeddingslong-document processingbatch inference

Trust & controls

deterministic validationhuman-in-the-loopprovenance & audit trailsmulti-tenant isolationpermission-aware access21 CFR Part 11 / GxP

Evals & observability

gold-set harnessesLLM-as-judgetoken cost & latency trackingmetricsstructured loggingtracing

Backend

PythonFastAPIFlaskDjangogRPC / RESTevent-driven servicesPostgreSQLMongoDBRedisKafkaCelery

Cloud & infra

GCP · Vertex AIAzure · AKSAWSDockerKubernetesTerraformSQLTypeScriptBash
Education

B.Tech, Chemical Engineering, Anna University (ACT Campus), Chennai · 2016–2020 · CGPA 8.19/10

Awards

4× Shining Star (monthly best performer), Saama · INSPIRE Science Camp, Govt. of India, top 200 nationally

Contact

Intelligence in the loop. Proof at the end.