FedLoRA-C: Communication-Efficient Federated Fine-Tuning of LLMs via SLC

Come join us on Oct. 20, 2026 (11:30 am - 1 pm) for our next IEEE CVT luncheon meeting at UTD.

Lokanatha Reddy Gandikota
(https://www.linkedin.com/in/lokanatha-reddy-gandikota/?isSelfProfile=false)

will present:

"FedLoRA-C: Communication-Efficient Federated Fine-Tuning of Large Language Models via Selective Layer Compression"

Join us each month if interested in great networking and thought-provoking meeting topics.

You do not need to be an IEEE member to attend.

Register at:
https://events.vtools.ieee.org/event/register/583543

(Feel free to use the "Register as a Guest" button if not a member of IEEE and be sure to click on the "I accept the IEEE Privacy Policy" box).

Meeting details and cost to attend are below
and at: https://events.vtools.ieee.org/m/583543

Reservation at IEEE site (vTools) is required by 4:30 pm Monday (10-19-26) if you'd like lunch. Free to attend if not having lunch.

On Oct. 20th:

This month's presenter will be:

Lokanatha Reddy Gandikota
(https://www.linkedin.com/in/lokanatha-reddy-gandikota/?isSelfProfile=false)

Abstract :

Fine-tuning large language models across distributed institutions — hospitals, law firms, financial organizations — is often blocked by two constraints: sensitive data cannot be centralized, and the communication cost of coordinating model updates becomes prohibitive at scale.

This talk presents FedLoRA-C, a communication-efficient framework for federated fine-tuning of LLMs.

By combining Low-Rank Adaptation with a novel Selective Layer Compression strategy that identifies and prioritizes the transformer layers most critical to task adaptation, FedLoRA-C reduces per-round communication cost by 67.3% while achieving accuracy within 1.4% of centralized fine-tuning.

The talk covers the core method, results across four NLP benchmarks including a healthcare question-answering task, and practical implications for deploying LLM adaptation in privacy-sensitive, bandwidth-constrained settings.

The work received the Best Industrial Paper Award at IEEE AIS 2026.

Bio:

Lokanatha Reddy Gandikota is a Distinguished Engineer at Verizon, Dallas, TX, with over 17 years of experience in software development, AI, and large-scale enterprise systems.

His recent work focuses on federated learning, LLM efficiency, and communication-efficient techniques for adapting large language models across distributed environments without centralizing sensitive data.

He has published research papers in IEEE and Springer venues and holds a pending patent through Verizon in the AI domain.

Lokanatha is an IEEE Senior Member and an IETE Fellow, actively engaged in bridging academic research with real-world industrial applications.

At this IEEE Dallas CVT Chapter meeting, he will present his work on FedLoRA-C, which received the Best Industrial Paper Award at IEEE AIS 2026.

Meeting Date: Oct. 20, 2026 (Program from 12 to 1 pm)

Location: UTD Synergy Park North Building; 3000 Waterview, Richardson, TX

Directions: NE Corner Synergy Park Blvd and Steward Dr.

Lunch Cost: see below.

Most importantly, come to network with fellow technologists, business execs, entrepreneurs and IEEE CVT members.
You do not have to be an IEEE member to attend.

Registration: Please pre-register at:
https://events.vtools.ieee.org/event/register/583543
(If you're not an IEEE member, you are welcome to use the "Register as Guest" button at the bottom of the page - be sure to click and check the "I accept the IEEE Privacy Policy" box).

Please pre-register by 4:30 pm on Monday (10-19-26) if you plan on having lunch.

Lunch will start at 11:30 am and the program runs from 12 to 1 pm.

You do not need to be a member of IEEE and the meeting is free if you do not intend to eat.

Location:
UTD Synergy Park North Building; 3000 Waterview, Richardson, TX
Directions: NE Corner of Synergy Park Blvd and Steward Dr. (map and more details below)

Prepay via the IEEE website by 4:30 pm Monday (10-19-26 ):
https://events.vtools.ieee.org/event/register/583543
IEEE Members $10.00
IEEE Life Members $5.00
Non-IEEE Members (Guests) $15.00

UTD students: $2.00 (First 5 Students who have registered in vTools as recorded within vTools date/time stamp on the registration will be given their $2 back at the event, provided they attend. All students must show UT Dallas ID at the desk the day of the meetings.)

Feel free to sign up using the "Register as Guest" button at the bottom of the page. You do not need to be an IEEE member to attend and the meeting is free to attend if you are not eating.

Pay in person the day of the meeting:
IEEE Members $15.00
IEEE Life Members $10.00
Non-IEEE Members (Guests) $20.00
UTD ECE students with ID $2.00

Anyone who has not pre-registered for lunch will be asked to wait until all pre-registered folks have had a chance to go through the buffet line.
Anyone not pre-registered will be charged the ‘Guest’ rate.

There is No charge for "No Lunch, just attending the talk".

Parking passes will be emailed to those that pre-register. Available at registration desk for those attending without registering.

Future meetings (subject to change but most likely):
Summer break = no meetings in June, July, Aug.
Nov. 17, 2026
Dec. 15, 2026
Jan. 19, 2027
Feb. 16, 2027
March 16, 2027
April 20, 2027
May 18, 2027