Tech At Bloomberg

We are engineers and designers, builders and makers delivering unrivaled products and services every day. We have a lot to share and a lot to learn. Join us.

[Tech At Bloomberg @TechAtBloomberg]

At the 2026 #EWFGlobal Annual Conference today (11:45 AM MT), Olayinka 'Rene' McQuick, a program manager in our CSOC, delivers a lightning talk about failing forward in one's #career and how to use adversity to one's advantage
https://t.co/p7DkJElKjS

[Joyee Cheung @JoyeeCheung]

It's been a great experience to contribute to the V8 upgrades for the 27 release and improve the V8 upgrade process in Node.js in the past months. Thanks @TechAtBloomberg for supporting my work on this :D

[Tech At Bloomberg @TechAtBloomberg]

In this morning's Day 3 Keynote at #SLDS2026 (9:00 AM ET), @Bloomberg’s David Rosenberg & Camilo Ortiz tackle bond price nowcasting - a problem so irregular it doesn’t fit neatly into any one statistical learning framework
https://t.co/sOku1Sm8rz
#Statistics #DataScience

[Tech At Bloomberg @TechAtBloomberg]

At #NxtAI2026 today (3:25 PM EST), @EnricoSantus of our CTO Office's AI Strategy & Research team, discusses 6 of his recent #FinancialAI research papers that explore how #AgenticAI & graph-based models help us move beyond simple text classification
https://t.co/XzJapTfaQw
#GenAI

[Tech At Bloomberg @TechAtBloomberg]

`slog` has been in @golang's standard library since 2023. Most projects barely use it. In today’s talk at #GoLab2026 (2:45 PM CET), software engineer Spencer Kormos shows the production patterns that make it useful, testable & maintainable
https://t.co/57dOibH4yJ
#golang

[Tech At Bloomberg @TechAtBloomberg]

@Columbia @BlackHatEvents Their research paper also states that these findings were disclosed to Google in September 2026, prior to the paper's publication
#reCAPTCHA
(4/4)

[Tech At Bloomberg @TechAtBloomberg]

At #ISC2026 today (9 AM CET; Session 7B), #softwaresecurity researcher Suphannee Sivakorn shares independent research showing that she broke Google’s #reCAPTCHA visual challenges 92.6% of the time (500 sessions) w/ free, #opensource, local #AI models
https://t.co/PL2zBuiuoU
(1/4)

[Tech At Bloomberg @TechAtBloomberg]

@Columbia @BlackHatEvents Read Sivakorn's new paper from Information Security Conference 2026 (#ISC2026), co-authored with #softwaresecurity researcher Samantha Gottlieb: "Robot Visions: Breaking #reCAPTCHA at Zero Cost and Zero Shot"
https://t.co/LRS5ZpqPoI
(3/4)

[Tech At Bloomberg @TechAtBloomberg]

Our CTO #DataScience Speaker Series welcomes @USCViterbi's Yue Zhao to our Engineering office in NY today to talk with our #AI researchers about the challenges of evaluating #agenticAI systems and the role of auditing in understanding them
https://t.co/kSd6tFhjEw
#AIEvaluation

[Tech At Bloomberg @TechAtBloomberg]

In her talk at #SHPE2026 today (9:30 AM ET), engineer Yi Jia makes the case that your #career is a canvas, not a conveyor belt - drawing on her own path through an #AI Ph.D., enterprise #ML & a deliberate pause to build a framework for non-linear growth
https://t.co/n5TocNQEXX

[Tech At Bloomberg @TechAtBloomberg]

@Bloomberg @UMassAmherst @NllpWorkshop @emnlpmeeting @EnricoSantus Read "ARGUS: Role-Aware Event Knowledge Graphs for U.S. Employment-Discrimination Complaints" by @UMassAmherst's Sriram Kannan, Swetha Saseendran & Vishnu Vardhan Reddy Kandi + our Leslie Barrett, Madhavan Seshadri & @EnricoSantus
https://t.co/EO0d5a0FpP
#NLLP2026 #LegalAI
(3/3)

[Tech At Bloomberg @TechAtBloomberg]

In today's #NLLP2026 Workshop (2:12 PM CET), @Bloomberg & @UMassAmherst #AI researchers present ARGUS, a pipeline that converts federal employment-discrimination complaints into Event #KnowledgeGraphs, capturing participant roles, temporal sequences & causal relations.
(1/3)

[Tech At Bloomberg @TechAtBloomberg]

@Bloomberg @UMassAmherst @NllpWorkshop @emnlpmeeting @EnricoSantus Legal cases often turn on *who did what to whom, and when* - not just keywords. ARGUS builds structured event graphs from U.S. employment-discrimination complaints, making legal cases comparable and searchable in ways text alone cannot.
#NLLP2026 #EMNLP2026 #LegalAI
(2/3)

[Tech At Bloomberg @TechAtBloomberg]

At #EMNLP2026 today (9:15 AM CET), Engineering Manager @daniel_preotiuc of our #AI Search Services team, delivers the Opening Keynote of the 11th Workshop on Financial Technology and Natural Language Processing, organized by SIG-FinTech
https://t.co/HiaDgCBddc
#FinNLP2026

[Tech At Bloomberg @TechAtBloomberg]

@debanjanbhucs MIDR teaches a retriever to understand them, without any serving-time image processing overhead. Instead, it serves retrieval via standard BM25F + dense search. Result: 23% gain over BM25 on ViDoRe V3, competitive with ColQwen2.5, at 9× smaller index size.
#EMNLP2026
(3/3)

[Tech At Bloomberg @TechAtBloomberg]

In Session 9 at #EMNLP2026, #AI researcher @debanjanbhucs presents MIDR, a new training-free framework that enriches documents with multimodal understanding at ingestion; it is faster, smaller + just as accurate.
#InformationRetrieval
(1/3)

[Tech At Bloomberg @TechAtBloomberg]

@debanjanbhucs Tables, charts & figures carry critical information that plain OCR loses. State-of-the-art retrieval serves every query over rendered page images - powerful, but slow & expensive. MIDR does multimodal reasoning at index time instead.
https://t.co/yAi8Ffzv9j
#EMNLP2026
(2/3)

[Tech At Bloomberg @TechAtBloomberg]

@jvanegue While GPT-5 & Claude Sonnet 4.5 can predict whether a program terminates nearly as well as top verification tool, ask them to *prove* it, and they break down. The researchers found a striking gap: models often know the answer, but can't show their work.
#EMNLP2026
(2/2)

[Tech At Bloomberg @TechAtBloomberg]

During Poster Session D at #EMNLP2026 today (11 AM CET), a group of researchers from Meta FAIR, UCL, ICL, HUJI, and Bloomberg, incl. @jvanegue, share research exposing a fundamental gap between #LLM semantic recognition and formal proof
https://t.co/IlWSIowLEG
(1/2)