Kalyan Venkatesh

Software Engineer

San Francisco, California, United States3 yrs 2 mos experience

AI EnabledAI ML Practitioner

Key Highlights

Led ML pipeline work reducing system latency by 40%.
Developed a three-agent system for hallucination monitoring.
Engineered data workflows improving responsiveness for 200+ users.

Stackforce AI infers this person is a Data Engineering and AI/ML specialist with a focus on educational technology and infrastructure management.

Contact

Skills

Core Skills

Machine LearningPythonData EngineeringData Management

Other Skills

Python (Programming Language)Pandas (Software)PostgreSQLWritten CommunicationNumPyDocker & KubernetesLangChainData PipelinesSoftware QualityData AnalyticsMLflowAnalyticsTechnology EducationHyperparameter OptimizationWeb Development

About

Software Engineer with 3+ years of experience building ML pipelines, data infrastructure & backend systems that keep ML systems reliable in production. My current graduate research investigates hallucination monitoring in agentic LLM pipelines. Can a lightweight runtime critic layer catch hallucinations without any model retraining? I built a three-agent system (Planner, Critic, Fixer) using LangGraph, Ollama and MLflow, and ran a full ablation study across 8 conditions on HumanEval. The honest result: the Critic works and it reliably detects hallucination risk. The Fixer, at 3B model scale, doesn't improve correctness and at the wrong threshold actively causes regression. That's the contribution - a clean empirical map of where lightweight runtime monitoring succeeds and where it breaks down. Phase 2 scales the Critic to 8B to find out if that changes things. Before grad school, I led data engineering and ML pipeline work at sensen.ai, cutting system latency by 40% and supporting model validation across 26 projects. At AECOM, I streamlined data integration pipelines for 100+ locations. Core stack: LangGraph · MLflow · Docker · Kubernetes · AWS (Certified Cloud Practitioner) · Python · TensorFlow · PyTorch · Hugging Face MS in Computer Science @ DePaul University (GPA 3.80, graduating June 2026). Available for full-time MLOps / AI/ML Engineer roles via OPT. Like to team up? Especially if you're working on model observability, LLM reliability, or agentic AI infrastructure, shoot me a message and we'll get going!

Experience

3 yrs 2 mos

Total Experience

1 yr 6 mos

Average Tenure

Current Experience

Depaul university

Graduate Research Engineer

Sep 2025 – Present · 8 mos · Chicago, Illinois, United States

Integrated LLM-based contextual assistants into student guidance prototypes for registration and case tracking, enabling real-time status resolution and guided task completion, estimated to reduce manual intervention by 25%.
Engineered backend and data workflows for academic systems including real-time scheduling updates, data preprocessing pipelines, and pub-sub-based announcement APIs, improving system responsiveness for 200+ daily users.
Conducted structured experimentation on runtime critic reliability in multi-agent LLM systems without model fine-tuning, executing temperature x threshold experiments across 400 evaluations to identify limitations of 3B-scale critic models.

Python (Programming Language)Machine LearningPython

Sensen.ai

Data Scientist

Aug 2022 – Aug 2024 · 2 yrs · Hyderabad, India

Initiated ANPR Accuracy/Discrepancy Analytics Pipeline through instance mapping across 26 projects, generating granular error insights, drift trends, and performance metrics for 5 model-quality KPIs to support model validation.
Developed end-to-end ETL pipelines for invoice and billing analytics, automating SQL extraction and transformation, generating structured reports based on user-defined parameters, nearly doubling processing efficiency.
Optimized high-volume data workflows by designing efficient PL/SQL functions and asynchronous image retrieval systems, delivering a 40% reduction in latency and enabling real-time enforcement data availability.
Managed end-to-end production deployment of SenFORCE modules including Parking Availability dashboards, 8 ticket generation APIs, and image processing pipelines; led cross-team issue resolution, reducing application downtime by 20%.
Led R&D on automated industrial pollution enforcement, integrating telemetry from drone GPS, sensors, and server-based ingestion pipelines to build a violation detection prototype projected to cut industrial contamination by 30–35%.

Pandas (Software)PostgreSQLData EngineeringMachine Learning

Aecom & siri

Data Engineer

May 2021 – Jul 2022 · 1 yr 2 mos · Chennai, India

Implemented data integration pipelines using Python (Pandas) & SQL for asset lifecycle management & capital planning across 25+ locations, enabling standardized analytics for 3+ cross-functional teams.
Built ETL pipelines to transform semi-structured data into standardized formats for operational dashboards & client-facing reports, consolidating data from project management systems to track 5+ metrics across active infrastructure engagements.

Written CommunicationData ManagementData Engineering