Venkata Keerthan N. profile photo 1 on Tal

Venkata Keerthan N.

Backend Engineer - Distributed Systems · Applied AI · Platform Engineering @ Adaequare | IITM

Hyderabad

2+ years of work experience

Adaequare
Mar 20266 mosPresent

Hyderabad, Telangana, India

I work on Surge Vector, a platform where business applications are defined in metadata instead of written as code. One entity definition drives the data model, the generated APIs, the workflows, and the screens. I built the workflow engine, which is my main piece of the platform. I also built several of the runtime pieces around it. Built the workflow execution layer on Temporal, including human-approval steps that can pause for days and resume correctly. Keeping workflow code deterministic is what makes Temporal replay it correctly after retries and restarts, so most of the care goes there. Wrote a runtime interpreter that walks the workflow graph directly with no compile step. A change made in the designer takes effect without a redeploy. Built the realtime GraphQL layer and the part of the metadata engine that holds entities and workflows in memory, keeping them current so the rest of the platform reads consistent state. Built the audit and logging layer, which came out of the workflow work. Durable execution is only useful if you can see exactly what ran and when. The engine integrates with the metadata, runtime, event/communication, and external API layers. Workflows are driven by metadata, triggered through the event layer, and call out through the API layer, so I had to understand each one well enough to build against it. We've built manufacturing, procurement, and sales ERP workflows on the platform, which is how we check whether the metadata model is general enough for real business domains.
QFI Research Capital
Sep 20253 mosNov 2025

Provide guidance on architecture, AI integration, and strategic planning for an algorithm-driven trade execution system. Advise on system architecture and future scalability, ensuring the platform supports growth and complex analytics. Contribute to strategic planning for AI integration, risk management improvements, and signal generation enhancements, guiding long-term development priorities.
QFI Research Capital
May 20254 mosAug 2025

Led initial development and strategic planning for an algorithm-driven trade execution system, focusing on real-time analytics, risk management, and signal generation. Built core backend and application components using PostgreSQL, InfluxDB, Docker, Kafka, Redis, Kite Connect, and integrated risk engines, enabling real-time trade execution, modular architecture, and automated data aggregation. Contributed to strategy for AI integration, risk modeling, and signal generation, laying the foundation for future portfolio management extensions and scalable system improvements.
Adaequare
Mar 20251 yrFeb 2026

Hyderabad, Telangana, India

Built document-intelligence and ETL systems for external clients, turning unstructured documents into structured data that downstream systems could use without manual re-entry. An OCR and LLM pipeline that extracts structured fields from property appraisal PDFs, over 10,000 documents a year, removing roughly 300 hours of manual data entry a month. Built the FastAPI backend and a Streamlit prototype for fast iteration (Tesseract, Detectron, Gemini, LangChain, React). A two-layer ETL system using heuristics and fuzzy/semantic matching with a model layer and automated validation, standardizing county tax-roll datasets of 20,000+ records each. It handles 50+ datasets a year and cut analyst review time by about 85%. The design scales to 3,000+ US counties. A contract-intelligence service that reads MSAs and other legal documents and produces structured JSON for CRM, ERP, and compliance workflows. Deployed on Cloud Run with Docker, async job processing, and API-key security. Across these, most of the work was in validation and handling the documents that didn't fit the expected pattern, including deciding when output wasn't reliable enough to pass through without a human check.
International and Alumni Relations Student Council, IIT Madras
Jan 20211 yr 1 moJan 2022

Chennai, Tamil Nadu, India

Addressed specific problems faced by the institute, particularly in the International and Alumni Relations sector. Analyzed current technological assets and applications utilized by IAR, identifying potential enhancements. Facilitated continuous improvement initiatives, encouraging innovation and excellence in operations.

Venkata’s AI Stack

antigravity
antigravity
codex
codex
cursor
cursor

Education

Indian Institute of Technology, Madras

Bachelor of Technology - BTech, Aerospace, Aeronautical and Astronautical Engineering

Indian Institute of Technology, Madras

20205 yrs2025

Join Venkata and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Venkata Keerthan N. on Tal

Headline

Backend Engineer - Distributed Systems · Applied AI · Platform Engineering @ Adaequare | IITM

About

Software Engineer building AI systems, workflow platforms, metadata-driven architectures, and business automation products. I love coding, music and travel.

Location

Hyderabad

AI stack

antigravity

codex

cursor

Work experience

Software Engineer - Applied AI & Platform Engineering at Adaequare, Mar 2026 - Present · 6 mos, Hyderabad, Telangana, India, I work on Surge Vector, a platform where business applications are defined in metadata instead of written as code. One entity definition drives the data model, the generated APIs, the workflows, and the screens. I built the workflow engine, which is my main piece of the platform. I also built several of the runtime pieces around it. Built the workflow execution layer on Temporal, including human-approval steps that can pause for days and resume correctly. Keeping workflow code deterministic is what makes Temporal replay it correctly after retries and restarts, so most of the care goes there. Wrote a runtime interpreter that walks the workflow graph directly with no compile step. A change made in the designer takes effect without a redeploy. Built the realtime GraphQL layer and the part of the metadata engine that holds entities and workflows in memory, keeping them current so the rest of the platform reads consistent state. Built the audit and logging layer, which came out of the workflow work. Durable execution is only useful if you can see exactly what ran and when. The engine integrates with the metadata, runtime, event/communication, and external API layers. Workflows are driven by metadata, triggered through the event layer, and call out through the API layer, so I had to understand each one well enough to build against it. We've built manufacturing, procurement, and sales ERP workflows on the platform, which is how we check whether the metadata model is general enough for real business domains.

Technical Strategy Consultant at QFI Research Capital, Sep 2025 - Nov 2025 · 3 mos, Provide guidance on architecture, AI integration, and strategic planning for an algorithm-driven trade execution system. Advise on system architecture and future scalability, ensuring the platform supports growth and complex analytics. Contribute to strategic planning for AI integration, risk management improvements, and signal generation enhancements, guiding long-term development priorities.

Founding Engineer at QFI Research Capital, May 2025 - Aug 2025 · 4 mos, Led initial development and strategic planning for an algorithm-driven trade execution system, focusing on real-time analytics, risk management, and signal generation. Built core backend and application components using PostgreSQL, InfluxDB, Docker, Kafka, Redis, Kite Connect, and integrated risk engines, enabling real-time trade execution, modular architecture, and automated data aggregation. Contributed to strategy for AI integration, risk modeling, and signal generation, laying the foundation for future portfolio management extensions and scalable system improvements.

Software Development Engineer - Applied AI at Adaequare, Mar 2025 - Feb 2026 · 1 yr, Hyderabad, Telangana, India, Built document-intelligence and ETL systems for external clients, turning unstructured documents into structured data that downstream systems could use without manual re-entry. An OCR and LLM pipeline that extracts structured fields from property appraisal PDFs, over 10,000 documents a year, removing roughly 300 hours of manual data entry a month. Built the FastAPI backend and a Streamlit prototype for fast iteration (Tesseract, Detectron, Gemini, LangChain, React). A two-layer ETL system using heuristics and fuzzy/semantic matching with a model layer and automated validation, standardizing county tax-roll datasets of 20,000+ records each. It handles 50+ datasets a year and cut analyst review time by about 85%. The design scales to 3,000+ US counties. A contract-intelligence service that reads MSAs and other legal documents and produces structured JSON for CRM, ERP, and compliance workflows. Deployed on Cloud Run with Docker, async job processing, and API-key security. Across these, most of the work was in validation and handling the documents that didn't fit the expected pattern, including deciding when output wasn't reliable enough to pass through without a human check.

Operations Manager at International and Alumni Relations Student Council, IIT Madras, Jan 2021 - Jan 2022 · 1 yr 1 mo, Chennai, Tamil Nadu, India, Addressed specific problems faced by the institute, particularly in the International and Alumni Relations sector. Analyzed current technological assets and applications utilized by IAR, identifying potential enhancements. Facilitated continuous improvement initiatives, encouraging innovation and excellence in operations.

Education

Bachelor of Technology - BTech, Aerospace, Aeronautical and Astronautical Engineering, Indian Institute of Technology, Madras